swift/test/unit/obj/test_diskfile.py

1669 lines
69 KiB
Python
Raw Normal View History

#-*- coding:utf-8 -*-
# Copyright (c) 2010-2012 OpenStack Foundation
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
# implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Tests for swift.obj.diskfile"""
import cPickle as pickle
import os
import errno
import mock
import unittest
import email
import tempfile
import xattr
from shutil import rmtree
from time import time
from tempfile import mkdtemp
from hashlib import md5
Object replication ssync (an rsync alternative) For this commit, ssync is just a direct replacement for how we use rsync. Assuming we switch over to ssync completely someday and drop rsync, we will then be able to improve the algorithms even further (removing local objects as we successfully transfer each one rather than waiting for whole partitions, using an index.db with hash-trees, etc., etc.) For easier review, this commit can be thought of in distinct parts: 1) New global_conf_callback functionality for allowing services to perform setup code before workers, etc. are launched. (This is then used by ssync in the object server to create a cross-worker semaphore to restrict concurrent incoming replication.) 2) A bit of shifting of items up from object server and replicator to diskfile or DEFAULT conf sections for better sharing of the same settings. conn_timeout, node_timeout, client_timeout, network_chunk_size, disk_chunk_size. 3) Modifications to the object server and replicator to optionally use ssync in place of rsync. This is done in a generic enough way that switching to FutureSync should be easy someday. 4) The biggest part, and (at least for now) completely optional part, are the new ssync_sender and ssync_receiver files. Nice and isolated for easier testing and visibility into test coverage, etc. All the usual logging, statsd, recon, etc. instrumentation is still there when using ssync, just as it is when using rsync. Beyond the essential error and exceptional condition logging, I have not added any additional instrumentation at this time. Unless there is something someone finds super pressing to have added to the logging, I think such additions would be better as separate change reviews. FOR NOW, IT IS NOT RECOMMENDED TO USE SSYNC ON PRODUCTION CLUSTERS. Some of us will be in a limited fashion to look for any subtle issues, tuning, etc. but generally ssync is an experimental feature. In its current implementation it is probably going to be a bit slower than rsync, but if all goes according to plan it will end up much faster. There are no comparisions yet between ssync and rsync other than some raw virtual machine testing I've done to show it should compete well enough once we can put it in use in the real world. If you Tweet, Google+, or whatever, be sure to indicate it's experimental. It'd be best to keep it out of deployment guides, howtos, etc. until we all figure out if we like it, find it to be stable, etc. Change-Id: If003dcc6f4109e2d2a42f4873a0779110fff16d6
2013-08-28 16:10:43 +00:00
from contextlib import closing, nested
from gzip import GzipFile
from eventlet import tpool
Alternate DiskFile constructor for efficient auditing. Before, to audit an object, the auditor: - calls listdir(object-hash-dir) - picks out the .data file from the listing - pulls out all N of its user.swift.metadata* xattrs - unpickles them - pulls out the value for 'name' - splits the name into a/c/o - then instantiates and opens a DiskFile(a, c, o), which does the following - joins a/c/o back into a name - hashes the name - calls listdir(object-hash-dir) (AGAIN) - picks out the .data file (and maybe .meta) from the listing (AGAIN) - pulls out all N of its user.swift.metadata* xattrs (AGAIN) - unpickles them (AGAIN) - starts reading object's contents off disk Now, the auditor simply locates the hash dir on the filesystem (saving one listdir) and then hands it off to DiskFileManager.get_diskfile_from_audit_location, which then instantiates a DiskFile in a way that lazy-loads the name later (saving one xattr reading). As part of this, DiskFile.open() will now quarantine a hash "directory" that's actually a file. Before, the audit location generator would skip those, but now they make it clear into DiskFile(). It's better to quarantine them anyway, as they're not doing any good the way they are. Also, removed the was_quarantined attribute on DiskFileReader. Now you can pass in a quarantine_hook callable to DiskFile.reader() that gets called if the file was quarantined. Default is to log quarantines, but otherwise do nothing. Change-Id: I04fc14569982a17fcc89e00832725ae71009335a
2013-10-28 14:57:18 -07:00
from test.unit import FakeLogger, mock as unit_mock, temptree
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
from swift.obj import diskfile
from swift.common import utils
from swift.common.utils import hash_path, mkdirs, normalize_timestamp
from swift.common import ring
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
from swift.common.exceptions import DiskFileNotExist, DiskFileQuarantined, \
DiskFileDeviceUnavailable, DiskFileDeleted, DiskFileNotOpen, \
DiskFileError, ReplicationLockTimeout
def _create_test_ring(path):
testgz = os.path.join(path, 'object.ring.gz')
intended_replica2part2dev_id = [
[0, 1, 2, 3, 4, 5, 6],
[1, 2, 3, 0, 5, 6, 4],
[2, 3, 0, 1, 6, 4, 5]]
intended_devs = [
{'id': 0, 'device': 'sda', 'zone': 0, 'ip': '127.0.0.0', 'port': 6000},
{'id': 1, 'device': 'sda', 'zone': 1, 'ip': '127.0.0.1', 'port': 6000},
{'id': 2, 'device': 'sda', 'zone': 2, 'ip': '127.0.0.2', 'port': 6000},
{'id': 3, 'device': 'sda', 'zone': 4, 'ip': '127.0.0.3', 'port': 6000},
{'id': 4, 'device': 'sda', 'zone': 5, 'ip': '127.0.0.4', 'port': 6000},
{'id': 5, 'device': 'sda', 'zone': 6,
'ip': 'fe80::202:b3ff:fe1e:8329', 'port': 6000},
{'id': 6, 'device': 'sda', 'zone': 7,
'ip': '2001:0db8:85a3:0000:0000:8a2e:0370:7334', 'port': 6000}]
intended_part_shift = 30
intended_reload_time = 15
with closing(GzipFile(testgz, 'wb')) as f:
pickle.dump(
ring.RingData(intended_replica2part2dev_id, intended_devs,
intended_part_shift),
f)
return ring.Ring(path, ring_name='object',
reload_time=intended_reload_time)
class TestDiskFileModuleMethods(unittest.TestCase):
def setUp(self):
utils.HASH_PATH_SUFFIX = 'endcap'
utils.HASH_PATH_PREFIX = ''
# Setup a test ring (stolen from common/test_ring.py)
self.testdir = tempfile.mkdtemp()
self.devices = os.path.join(self.testdir, 'node')
rmtree(self.testdir, ignore_errors=1)
os.mkdir(self.testdir)
os.mkdir(self.devices)
os.mkdir(os.path.join(self.devices, 'sda'))
self.objects = os.path.join(self.devices, 'sda', 'objects')
os.mkdir(self.objects)
self.parts = {}
for part in ['0', '1', '2', '3']:
self.parts[part] = os.path.join(self.objects, part)
os.mkdir(os.path.join(self.objects, part))
self.ring = _create_test_ring(self.testdir)
self.conf = dict(
swift_dir=self.testdir, devices=self.devices, mount_check='false',
timeout='300', stats_interval='1')
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
self.df_mgr = diskfile.DiskFileManager(self.conf, FakeLogger())
def tearDown(self):
rmtree(self.testdir, ignore_errors=1)
def test_hash_suffix_hash_dir_is_file_quarantine(self):
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda', '0', 'a', 'c', 'o')
mkdirs(os.path.dirname(df._datadir))
open(df._datadir, 'wb').close()
ohash = hash_path('a', 'c', 'o')
data_dir = ohash[-3:]
whole_path_from = os.path.join(self.objects, '0', data_dir)
orig_quarantine_renamer = diskfile.quarantine_renamer
called = [False]
def wrapped(*args, **kwargs):
called[0] = True
return orig_quarantine_renamer(*args, **kwargs)
try:
diskfile.quarantine_renamer = wrapped
diskfile.hash_suffix(whole_path_from, 101)
finally:
diskfile.quarantine_renamer = orig_quarantine_renamer
self.assertTrue(called[0])
def test_hash_suffix_one_file(self):
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda', '0', 'a', 'c', 'o')
mkdirs(df._datadir)
f = open(
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
os.path.join(df._datadir,
normalize_timestamp(time() - 100) + '.ts'),
'wb')
f.write('1234567890')
f.close()
ohash = hash_path('a', 'c', 'o')
data_dir = ohash[-3:]
whole_path_from = os.path.join(self.objects, '0', data_dir)
diskfile.hash_suffix(whole_path_from, 101)
self.assertEquals(len(os.listdir(self.parts['0'])), 1)
diskfile.hash_suffix(whole_path_from, 99)
self.assertEquals(len(os.listdir(self.parts['0'])), 0)
def test_hash_suffix_multi_file_one(self):
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda', '0', 'a', 'c', 'o')
mkdirs(df._datadir)
for tdiff in [1, 50, 100, 500]:
for suff in ['.meta', '.data', '.ts']:
f = open(
os.path.join(
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df._datadir,
normalize_timestamp(int(time()) - tdiff) + suff),
'wb')
f.write('1234567890')
f.close()
ohash = hash_path('a', 'c', 'o')
data_dir = ohash[-3:]
whole_path_from = os.path.join(self.objects, '0', data_dir)
hsh_path = os.listdir(whole_path_from)[0]
whole_hsh_path = os.path.join(whole_path_from, hsh_path)
diskfile.hash_suffix(whole_path_from, 99)
# only the tombstone should be left
self.assertEquals(len(os.listdir(whole_hsh_path)), 1)
def test_hash_suffix_multi_file_two(self):
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda', '0', 'a', 'c', 'o')
mkdirs(df._datadir)
for tdiff in [1, 50, 100, 500]:
suffs = ['.meta', '.data']
if tdiff > 50:
suffs.append('.ts')
for suff in suffs:
f = open(
os.path.join(
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df._datadir,
normalize_timestamp(int(time()) - tdiff) + suff),
'wb')
f.write('1234567890')
f.close()
ohash = hash_path('a', 'c', 'o')
data_dir = ohash[-3:]
whole_path_from = os.path.join(self.objects, '0', data_dir)
hsh_path = os.listdir(whole_path_from)[0]
whole_hsh_path = os.path.join(whole_path_from, hsh_path)
diskfile.hash_suffix(whole_path_from, 99)
# only the meta and data should be left
self.assertEquals(len(os.listdir(whole_hsh_path)), 2)
def test_invalidate_hash(self):
def assertFileData(file_path, data):
with open(file_path, 'r') as fp:
fdata = fp.read()
self.assertEquals(pickle.loads(fdata), pickle.loads(data))
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda', '0', 'a', 'c', 'o')
mkdirs(df._datadir)
ohash = hash_path('a', 'c', 'o')
data_dir = ohash[-3:]
whole_path_from = os.path.join(self.objects, '0', data_dir)
hashes_file = os.path.join(self.objects, '0',
diskfile.HASH_FILE)
# test that non existent file except caught
self.assertEquals(diskfile.invalidate_hash(whole_path_from),
None)
# test that hashes get cleared
check_pickle_data = pickle.dumps({data_dir: None},
diskfile.PICKLE_PROTOCOL)
for data_hash in [{data_dir: None}, {data_dir: 'abcdefg'}]:
with open(hashes_file, 'wb') as fp:
pickle.dump(data_hash, fp, diskfile.PICKLE_PROTOCOL)
diskfile.invalidate_hash(whole_path_from)
assertFileData(hashes_file, check_pickle_data)
def test_get_hashes(self):
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda', '0', 'a', 'c', 'o')
mkdirs(df._datadir)
with open(
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
os.path.join(df._datadir,
normalize_timestamp(time()) + '.ts'),
'wb') as f:
f.write('1234567890')
part = os.path.join(self.objects, '0')
hashed, hashes = diskfile.get_hashes(part)
self.assertEquals(hashed, 1)
self.assert_('a83' in hashes)
hashed, hashes = diskfile.get_hashes(part, do_listdir=True)
self.assertEquals(hashed, 0)
self.assert_('a83' in hashes)
hashed, hashes = diskfile.get_hashes(part, recalculate=['a83'])
self.assertEquals(hashed, 1)
self.assert_('a83' in hashes)
def test_get_hashes_bad_dir(self):
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda', '0', 'a', 'c', 'o')
mkdirs(df._datadir)
with open(os.path.join(self.objects, '0', 'bad'), 'wb') as f:
f.write('1234567890')
part = os.path.join(self.objects, '0')
hashed, hashes = diskfile.get_hashes(part)
self.assertEquals(hashed, 1)
self.assert_('a83' in hashes)
self.assert_('bad' not in hashes)
def test_get_hashes_unmodified(self):
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda', '0', 'a', 'c', 'o')
mkdirs(df._datadir)
with open(
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
os.path.join(df._datadir,
normalize_timestamp(time()) + '.ts'),
'wb') as f:
f.write('1234567890')
part = os.path.join(self.objects, '0')
hashed, hashes = diskfile.get_hashes(part)
i = [0]
def _getmtime(filename):
i[0] += 1
return 1
with unit_mock({'swift.obj.diskfile.getmtime': _getmtime}):
hashed, hashes = diskfile.get_hashes(
part, recalculate=['a83'])
self.assertEquals(i[0], 2)
def test_get_hashes_unmodified_and_zero_bytes(self):
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda', '0', 'a', 'c', 'o')
mkdirs(df._datadir)
part = os.path.join(self.objects, '0')
open(os.path.join(part, diskfile.HASH_FILE), 'w')
# Now the hash file is zero bytes.
i = [0]
def _getmtime(filename):
i[0] += 1
return 1
with unit_mock({'swift.obj.diskfile.getmtime': _getmtime}):
hashed, hashes = diskfile.get_hashes(
part, recalculate=[])
# getmtime will actually not get called. Initially, the pickle.load
# will raise an exception first and later, force_rewrite will
# short-circuit the if clause to determine whether to write out a
# fresh hashes_file.
self.assertEquals(i[0], 0)
self.assertTrue('a83' in hashes)
def test_get_hashes_modified(self):
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda', '0', 'a', 'c', 'o')
mkdirs(df._datadir)
with open(
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
os.path.join(df._datadir,
normalize_timestamp(time()) + '.ts'),
'wb') as f:
f.write('1234567890')
part = os.path.join(self.objects, '0')
hashed, hashes = diskfile.get_hashes(part)
i = [0]
def _getmtime(filename):
if i[0] < 3:
i[0] += 1
return i[0]
with unit_mock({'swift.obj.diskfile.getmtime': _getmtime}):
hashed, hashes = diskfile.get_hashes(
part, recalculate=['a83'])
self.assertEquals(i[0], 3)
def check_hash_cleanup_listdir(self, input_files, output_files):
file_list = list(input_files)
def mock_listdir(path):
return list(file_list)
def mock_unlink(path):
file_list.remove(os.path.basename(path))
with unit_mock({'os.listdir': mock_listdir, 'os.unlink': mock_unlink}):
self.assertEquals(diskfile.hash_cleanup_listdir('/whatever'),
output_files)
def test_hash_cleanup_listdir_purge_data_newer_ts(self):
# purge .data if there's a newer .ts
file1 = normalize_timestamp(time()) + '.data'
file2 = normalize_timestamp(time() + 1) + '.ts'
file_list = [file1, file2]
self.check_hash_cleanup_listdir(file_list, [file2])
def test_hash_cleanup_listdir_purge_ts_newer_data(self):
# purge .ts if there's a newer .data
file1 = normalize_timestamp(time()) + '.ts'
file2 = normalize_timestamp(time() + 1) + '.data'
file_list = [file1, file2]
self.check_hash_cleanup_listdir(file_list, [file2])
def test_hash_cleanup_listdir_keep_meta_data_purge_ts(self):
# keep .meta and .data if meta newer than data and purge .ts
file1 = normalize_timestamp(time()) + '.ts'
file2 = normalize_timestamp(time() + 1) + '.data'
file3 = normalize_timestamp(time() + 2) + '.meta'
file_list = [file1, file2, file3]
self.check_hash_cleanup_listdir(file_list, [file3, file2])
def test_hash_cleanup_listdir_keep_one_ts(self):
# keep only latest of multiple .ts files
file1 = normalize_timestamp(time()) + '.ts'
file2 = normalize_timestamp(time() + 1) + '.ts'
file3 = normalize_timestamp(time() + 2) + '.ts'
file_list = [file1, file2, file3]
self.check_hash_cleanup_listdir(file_list, [file3])
def test_hash_cleanup_listdir_keep_one_data(self):
# keep only latest of multiple .data files
file1 = normalize_timestamp(time()) + '.data'
file2 = normalize_timestamp(time() + 1) + '.data'
file3 = normalize_timestamp(time() + 2) + '.data'
file_list = [file1, file2, file3]
self.check_hash_cleanup_listdir(file_list, [file3])
def test_hash_cleanup_listdir_keep_one_meta(self):
# keep only latest of multiple .meta files
file1 = normalize_timestamp(time()) + '.data'
file2 = normalize_timestamp(time() + 1) + '.meta'
file3 = normalize_timestamp(time() + 2) + '.meta'
file_list = [file1, file2, file3]
self.check_hash_cleanup_listdir(file_list, [file3, file1])
def test_hash_cleanup_listdir_ignore_orphaned_ts(self):
# A more recent orphaned .meta file will prevent old .ts files
# from being cleaned up otherwise
file1 = normalize_timestamp(time()) + '.ts'
file2 = normalize_timestamp(time() + 1) + '.ts'
file3 = normalize_timestamp(time() + 2) + '.meta'
file_list = [file1, file2, file3]
self.check_hash_cleanup_listdir(file_list, [file3, file2])
def test_hash_cleanup_listdir_purge_old_data_only(self):
# Oldest .data will be purge, .meta and .ts won't be touched
file1 = normalize_timestamp(time()) + '.data'
file2 = normalize_timestamp(time() + 1) + '.ts'
file3 = normalize_timestamp(time() + 2) + '.meta'
file_list = [file1, file2, file3]
self.check_hash_cleanup_listdir(file_list, [file3, file2])
def test_hash_cleanup_listdir_purge_old_ts(self):
# A single old .ts file will be removed
file1 = normalize_timestamp(time() - (diskfile.ONE_WEEK + 1)) + '.ts'
file_list = [file1]
self.check_hash_cleanup_listdir(file_list, [])
def test_hash_cleanup_listdir_meta_keeps_old_ts(self):
# An orphaned .meta will not clean up a very old .ts
file1 = normalize_timestamp(time() - (diskfile.ONE_WEEK + 1)) + '.ts'
file2 = normalize_timestamp(time() + 2) + '.meta'
file_list = [file1, file2]
self.check_hash_cleanup_listdir(file_list, [file2, file1])
def test_hash_cleanup_listdir_keep_single_old_data(self):
# A single old .data file will not be removed
file1 = normalize_timestamp(time() - (diskfile.ONE_WEEK + 1)) + '.data'
file_list = [file1]
self.check_hash_cleanup_listdir(file_list, [file1])
def test_hash_cleanup_listdir_keep_single_old_meta(self):
# A single old .meta file will not be removed
file1 = normalize_timestamp(time() - (diskfile.ONE_WEEK + 1)) + '.meta'
file_list = [file1]
self.check_hash_cleanup_listdir(file_list, [file1])
Alternate DiskFile constructor for efficient auditing. Before, to audit an object, the auditor: - calls listdir(object-hash-dir) - picks out the .data file from the listing - pulls out all N of its user.swift.metadata* xattrs - unpickles them - pulls out the value for 'name' - splits the name into a/c/o - then instantiates and opens a DiskFile(a, c, o), which does the following - joins a/c/o back into a name - hashes the name - calls listdir(object-hash-dir) (AGAIN) - picks out the .data file (and maybe .meta) from the listing (AGAIN) - pulls out all N of its user.swift.metadata* xattrs (AGAIN) - unpickles them (AGAIN) - starts reading object's contents off disk Now, the auditor simply locates the hash dir on the filesystem (saving one listdir) and then hands it off to DiskFileManager.get_diskfile_from_audit_location, which then instantiates a DiskFile in a way that lazy-loads the name later (saving one xattr reading). As part of this, DiskFile.open() will now quarantine a hash "directory" that's actually a file. Before, the audit location generator would skip those, but now they make it clear into DiskFile(). It's better to quarantine them anyway, as they're not doing any good the way they are. Also, removed the was_quarantined attribute on DiskFileReader. Now you can pass in a quarantine_hook callable to DiskFile.reader() that gets called if the file was quarantined. Default is to log quarantines, but otherwise do nothing. Change-Id: I04fc14569982a17fcc89e00832725ae71009335a
2013-10-28 14:57:18 -07:00
class TestObjectAuditLocationGenerator(unittest.TestCase):
def _make_file(self, path):
try:
os.makedirs(os.path.dirname(path))
except OSError as err:
if err.errno != errno.EEXIST:
raise
with open(path, 'w'):
pass
def test_finding_of_hashdirs(self):
with temptree([]) as tmpdir:
# the good
os.makedirs(os.path.join(tmpdir, "sdp", "objects", "1519", "aca",
"5c1fdc1ffb12e5eaf84edc30d8b67aca"))
os.makedirs(os.path.join(tmpdir, "sdp", "objects", "1519", "aca",
"fdfd184d39080020bc8b487f8a7beaca"))
os.makedirs(os.path.join(tmpdir, "sdp", "objects", "1519", "df2",
"b0fe7af831cc7b1af5bf486b1c841df2"))
os.makedirs(os.path.join(tmpdir, "sdp", "objects", "9720", "ca5",
"4a943bc72c2e647c4675923d58cf4ca5"))
os.makedirs(os.path.join(tmpdir, "sdq", "objects", "3071", "8eb",
"fcd938702024c25fef6c32fef05298eb"))
# the bad
self._make_file(os.path.join(tmpdir, "sdp", "objects", "1519",
"fed"))
self._make_file(os.path.join(tmpdir, "sdq", "objects", "9876"))
# the empty
os.makedirs(os.path.join(tmpdir, "sdr"))
os.makedirs(os.path.join(tmpdir, "sds", "objects"))
os.makedirs(os.path.join(tmpdir, "sdt", "objects", "9601"))
os.makedirs(os.path.join(tmpdir, "sdu", "objects", "6499", "f80"))
# the irrelevant
os.makedirs(os.path.join(tmpdir, "sdv", "accounts", "77", "421",
"4b8c86149a6d532f4af018578fd9f421"))
os.makedirs(os.path.join(tmpdir, "sdw", "containers", "28", "51e",
"4f9eee668b66c6f0250bfa3c7ab9e51e"))
locations = [(loc.path, loc.device, loc.partition)
for loc in diskfile.object_audit_location_generator(
devices=tmpdir, mount_check=False)]
locations.sort()
self.assertEqual(
locations,
[(os.path.join(tmpdir, "sdp", "objects", "1519", "aca",
"5c1fdc1ffb12e5eaf84edc30d8b67aca"),
"sdp", "1519"),
(os.path.join(tmpdir, "sdp", "objects", "1519", "aca",
"fdfd184d39080020bc8b487f8a7beaca"),
"sdp", "1519"),
(os.path.join(tmpdir, "sdp", "objects", "1519", "df2",
"b0fe7af831cc7b1af5bf486b1c841df2"),
"sdp", "1519"),
(os.path.join(tmpdir, "sdp", "objects", "9720", "ca5",
"4a943bc72c2e647c4675923d58cf4ca5"),
"sdp", "9720"),
(os.path.join(tmpdir, "sdq", "objects", "3071", "8eb",
"fcd938702024c25fef6c32fef05298eb"),
"sdq", "3071")])
def test_skipping_unmounted_devices(self):
def mock_ismount(path):
return path.endswith('sdp')
with mock.patch('swift.obj.diskfile.ismount', mock_ismount):
Alternate DiskFile constructor for efficient auditing. Before, to audit an object, the auditor: - calls listdir(object-hash-dir) - picks out the .data file from the listing - pulls out all N of its user.swift.metadata* xattrs - unpickles them - pulls out the value for 'name' - splits the name into a/c/o - then instantiates and opens a DiskFile(a, c, o), which does the following - joins a/c/o back into a name - hashes the name - calls listdir(object-hash-dir) (AGAIN) - picks out the .data file (and maybe .meta) from the listing (AGAIN) - pulls out all N of its user.swift.metadata* xattrs (AGAIN) - unpickles them (AGAIN) - starts reading object's contents off disk Now, the auditor simply locates the hash dir on the filesystem (saving one listdir) and then hands it off to DiskFileManager.get_diskfile_from_audit_location, which then instantiates a DiskFile in a way that lazy-loads the name later (saving one xattr reading). As part of this, DiskFile.open() will now quarantine a hash "directory" that's actually a file. Before, the audit location generator would skip those, but now they make it clear into DiskFile(). It's better to quarantine them anyway, as they're not doing any good the way they are. Also, removed the was_quarantined attribute on DiskFileReader. Now you can pass in a quarantine_hook callable to DiskFile.reader() that gets called if the file was quarantined. Default is to log quarantines, but otherwise do nothing. Change-Id: I04fc14569982a17fcc89e00832725ae71009335a
2013-10-28 14:57:18 -07:00
with temptree([]) as tmpdir:
os.makedirs(os.path.join(tmpdir, "sdp", "objects",
"2607", "df3",
"ec2871fe724411f91787462f97d30df3"))
os.makedirs(os.path.join(tmpdir, "sdq", "objects",
"9785", "a10",
"4993d582f41be9771505a8d4cb237a10"))
locations = [
(loc.path, loc.device, loc.partition)
for loc in diskfile.object_audit_location_generator(
devices=tmpdir, mount_check=True)]
locations.sort()
self.assertEqual(
locations,
[(os.path.join(tmpdir, "sdp", "objects",
"2607", "df3",
"ec2871fe724411f91787462f97d30df3"),
"sdp", "2607")])
def test_only_catch_expected_errors(self):
# Crazy exceptions should still escape object_audit_location_generator
# so that errors get logged and a human can see what's going wrong;
# only normal FS corruption should be skipped over silently.
def list_locations(dirname):
return [(loc.path, loc.device, loc.partition)
for loc in diskfile.object_audit_location_generator(
devices=dirname, mount_check=False)]
real_listdir = os.listdir
def splode_if_endswith(suffix):
def sploder(path):
if path.endswith(suffix):
raise OSError(errno.EACCES, "don't try to ad-lib")
Alternate DiskFile constructor for efficient auditing. Before, to audit an object, the auditor: - calls listdir(object-hash-dir) - picks out the .data file from the listing - pulls out all N of its user.swift.metadata* xattrs - unpickles them - pulls out the value for 'name' - splits the name into a/c/o - then instantiates and opens a DiskFile(a, c, o), which does the following - joins a/c/o back into a name - hashes the name - calls listdir(object-hash-dir) (AGAIN) - picks out the .data file (and maybe .meta) from the listing (AGAIN) - pulls out all N of its user.swift.metadata* xattrs (AGAIN) - unpickles them (AGAIN) - starts reading object's contents off disk Now, the auditor simply locates the hash dir on the filesystem (saving one listdir) and then hands it off to DiskFileManager.get_diskfile_from_audit_location, which then instantiates a DiskFile in a way that lazy-loads the name later (saving one xattr reading). As part of this, DiskFile.open() will now quarantine a hash "directory" that's actually a file. Before, the audit location generator would skip those, but now they make it clear into DiskFile(). It's better to quarantine them anyway, as they're not doing any good the way they are. Also, removed the was_quarantined attribute on DiskFileReader. Now you can pass in a quarantine_hook callable to DiskFile.reader() that gets called if the file was quarantined. Default is to log quarantines, but otherwise do nothing. Change-Id: I04fc14569982a17fcc89e00832725ae71009335a
2013-10-28 14:57:18 -07:00
else:
return real_listdir(path)
return sploder
with temptree([]) as tmpdir:
os.makedirs(os.path.join(tmpdir, "sdf", "objects",
"2607", "b54",
"fe450ec990a88cc4b252b181bab04b54"))
with mock.patch('os.listdir', splode_if_endswith("sdf/objects")):
self.assertRaises(OSError, list_locations, tmpdir)
with mock.patch('os.listdir', splode_if_endswith("2607")):
self.assertRaises(OSError, list_locations, tmpdir)
with mock.patch('os.listdir', splode_if_endswith("b54")):
self.assertRaises(OSError, list_locations, tmpdir)
class TestDiskFileManager(unittest.TestCase):
def setUp(self):
self.tmpdir = mkdtemp()
self.testdir = os.path.join(
self.tmpdir, 'tmp_test_obj_server_DiskFile')
mkdirs(os.path.join(self.testdir, 'sda1', 'tmp'))
mkdirs(os.path.join(self.testdir, 'sda2', 'tmp'))
self._orig_tpool_exc = tpool.execute
tpool.execute = lambda f, *args, **kwargs: f(*args, **kwargs)
self.conf = dict(devices=self.testdir, mount_check='false',
keep_cache_size=2 * 1024)
self.df_mgr = diskfile.DiskFileManager(self.conf, FakeLogger())
def tearDown(self):
rmtree(self.tmpdir, ignore_errors=1)
def test_replication_lock_on(self):
# Double check settings
self.df_mgr.replication_one_per_device = True
self.df_mgr.replication_lock_timeout = 0.1
dev_path = os.path.join(self.testdir, 'sda1')
with self.df_mgr.replication_lock(dev_path):
lock_exc = None
exc = None
try:
with self.df_mgr.replication_lock(dev_path):
raise Exception(
'%r was not replication locked!' % dev_path)
except ReplicationLockTimeout as err:
lock_exc = err
except Exception as err:
exc = err
self.assertTrue(lock_exc is not None)
self.assertTrue(exc is None)
def test_replication_lock_off(self):
# Double check settings
self.df_mgr.replication_one_per_device = False
self.df_mgr.replication_lock_timeout = 0.1
dev_path = os.path.join(self.testdir, 'sda1')
with self.df_mgr.replication_lock(dev_path):
lock_exc = None
exc = None
try:
with self.df_mgr.replication_lock(dev_path):
raise Exception(
'%r was not replication locked!' % dev_path)
except ReplicationLockTimeout as err:
lock_exc = err
except Exception as err:
exc = err
self.assertTrue(lock_exc is None)
self.assertTrue(exc is not None)
def test_replication_lock_another_device_fine(self):
# Double check settings
self.df_mgr.replication_one_per_device = True
self.df_mgr.replication_lock_timeout = 0.1
dev_path = os.path.join(self.testdir, 'sda1')
dev_path2 = os.path.join(self.testdir, 'sda2')
with self.df_mgr.replication_lock(dev_path):
lock_exc = None
try:
with self.df_mgr.replication_lock(dev_path2):
pass
except ReplicationLockTimeout as err:
lock_exc = err
self.assertTrue(lock_exc is None)
class TestDiskFile(unittest.TestCase):
"""Test swift.obj.diskfile.DiskFile"""
def setUp(self):
"""Set up for testing swift.obj.diskfile"""
self.tmpdir = mkdtemp()
self.testdir = os.path.join(
self.tmpdir, 'tmp_test_obj_server_DiskFile')
mkdirs(os.path.join(self.testdir, 'sda1', 'tmp'))
self._orig_tpool_exc = tpool.execute
tpool.execute = lambda f, *args, **kwargs: f(*args, **kwargs)
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
self.conf = dict(devices=self.testdir, mount_check='false',
keep_cache_size=2 * 1024)
self.df_mgr = diskfile.DiskFileManager(self.conf, FakeLogger())
def tearDown(self):
"""Tear down for testing swift.obj.diskfile"""
rmtree(self.tmpdir, ignore_errors=1)
tpool.execute = self._orig_tpool_exc
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
def _create_ondisk_file(self, df, data, timestamp, metadata=None,
ext='.data'):
mkdirs(df._datadir)
if timestamp is None:
timestamp = time()
timestamp = normalize_timestamp(timestamp)
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
if not metadata:
metadata = {}
if 'X-Timestamp' not in metadata:
metadata['X-Timestamp'] = normalize_timestamp(timestamp)
if 'ETag' not in metadata:
etag = md5()
etag.update(data)
metadata['ETag'] = etag.hexdigest()
if 'name' not in metadata:
metadata['name'] = '/a/c/o'
if 'Content-Length' not in metadata:
metadata['Content-Length'] = str(len(data))
data_file = os.path.join(df._datadir, timestamp + ext)
with open(data_file, 'wb') as f:
f.write(data)
xattr.setxattr(f.fileno(), diskfile.METADATA_KEY,
pickle.dumps(metadata, diskfile.PICKLE_PROTOCOL))
Alternate DiskFile constructor for efficient auditing. Before, to audit an object, the auditor: - calls listdir(object-hash-dir) - picks out the .data file from the listing - pulls out all N of its user.swift.metadata* xattrs - unpickles them - pulls out the value for 'name' - splits the name into a/c/o - then instantiates and opens a DiskFile(a, c, o), which does the following - joins a/c/o back into a name - hashes the name - calls listdir(object-hash-dir) (AGAIN) - picks out the .data file (and maybe .meta) from the listing (AGAIN) - pulls out all N of its user.swift.metadata* xattrs (AGAIN) - unpickles them (AGAIN) - starts reading object's contents off disk Now, the auditor simply locates the hash dir on the filesystem (saving one listdir) and then hands it off to DiskFileManager.get_diskfile_from_audit_location, which then instantiates a DiskFile in a way that lazy-loads the name later (saving one xattr reading). As part of this, DiskFile.open() will now quarantine a hash "directory" that's actually a file. Before, the audit location generator would skip those, but now they make it clear into DiskFile(). It's better to quarantine them anyway, as they're not doing any good the way they are. Also, removed the was_quarantined attribute on DiskFileReader. Now you can pass in a quarantine_hook callable to DiskFile.reader() that gets called if the file was quarantined. Default is to log quarantines, but otherwise do nothing. Change-Id: I04fc14569982a17fcc89e00832725ae71009335a
2013-10-28 14:57:18 -07:00
def _create_test_file(self, data, timestamp=None, metadata=None,
account='a', container='c', object='o'):
if metadata is None:
metadata = {}
metadata.setdefault('name', '/%s/%s/%s' % (account, container, object))
df = self.df_mgr.get_diskfile('sda', '0', account, container, object)
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
self._create_ondisk_file(df, data, timestamp, metadata)
Alternate DiskFile constructor for efficient auditing. Before, to audit an object, the auditor: - calls listdir(object-hash-dir) - picks out the .data file from the listing - pulls out all N of its user.swift.metadata* xattrs - unpickles them - pulls out the value for 'name' - splits the name into a/c/o - then instantiates and opens a DiskFile(a, c, o), which does the following - joins a/c/o back into a name - hashes the name - calls listdir(object-hash-dir) (AGAIN) - picks out the .data file (and maybe .meta) from the listing (AGAIN) - pulls out all N of its user.swift.metadata* xattrs (AGAIN) - unpickles them (AGAIN) - starts reading object's contents off disk Now, the auditor simply locates the hash dir on the filesystem (saving one listdir) and then hands it off to DiskFileManager.get_diskfile_from_audit_location, which then instantiates a DiskFile in a way that lazy-loads the name later (saving one xattr reading). As part of this, DiskFile.open() will now quarantine a hash "directory" that's actually a file. Before, the audit location generator would skip those, but now they make it clear into DiskFile(). It's better to quarantine them anyway, as they're not doing any good the way they are. Also, removed the was_quarantined attribute on DiskFileReader. Now you can pass in a quarantine_hook callable to DiskFile.reader() that gets called if the file was quarantined. Default is to log quarantines, but otherwise do nothing. Change-Id: I04fc14569982a17fcc89e00832725ae71009335a
2013-10-28 14:57:18 -07:00
df = self.df_mgr.get_diskfile('sda', '0', account, container, object)
df.open()
return df
def test_get_metadata(self):
df = self._create_test_file('1234567890', timestamp=42)
md = df.get_metadata()
self.assertEquals(md['X-Timestamp'], normalize_timestamp(42))
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
def test_get_metadata_not_opened(self):
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
self.assertRaises(DiskFileNotOpen, df.get_metadata)
def test_not_opened(self):
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
try:
with df:
pass
except DiskFileNotOpen:
pass
else:
self.fail("Expected DiskFileNotOpen exception")
def test_disk_file_default_disallowed_metadata(self):
# build an object with some meta (ts 41)
orig_metadata = {'X-Object-Meta-Key1': 'Value1',
'Content-Type': 'text/garbage'}
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self._get_open_disk_file(ts=41, extra_metadata=orig_metadata)
with df.open():
self.assertEquals('1024', df._metadata['Content-Length'])
# write some new metadata (fast POST, don't send orig meta, ts 42)
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
df.write_metadata({'X-Timestamp': normalize_timestamp(42),
'X-Object-Meta-Key2': 'Value2'})
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
with df.open():
# non-fast-post updateable keys are preserved
self.assertEquals('text/garbage', df._metadata['Content-Type'])
# original fast-post updateable keys are removed
self.assert_('X-Object-Meta-Key1' not in df._metadata)
# new fast-post updateable keys are added
self.assertEquals('Value2', df._metadata['X-Object-Meta-Key2'])
def test_disk_file_app_iter_corners(self):
df = self._create_test_file('1234567890')
Alternate DiskFile constructor for efficient auditing. Before, to audit an object, the auditor: - calls listdir(object-hash-dir) - picks out the .data file from the listing - pulls out all N of its user.swift.metadata* xattrs - unpickles them - pulls out the value for 'name' - splits the name into a/c/o - then instantiates and opens a DiskFile(a, c, o), which does the following - joins a/c/o back into a name - hashes the name - calls listdir(object-hash-dir) (AGAIN) - picks out the .data file (and maybe .meta) from the listing (AGAIN) - pulls out all N of its user.swift.metadata* xattrs (AGAIN) - unpickles them (AGAIN) - starts reading object's contents off disk Now, the auditor simply locates the hash dir on the filesystem (saving one listdir) and then hands it off to DiskFileManager.get_diskfile_from_audit_location, which then instantiates a DiskFile in a way that lazy-loads the name later (saving one xattr reading). As part of this, DiskFile.open() will now quarantine a hash "directory" that's actually a file. Before, the audit location generator would skip those, but now they make it clear into DiskFile(). It's better to quarantine them anyway, as they're not doing any good the way they are. Also, removed the was_quarantined attribute on DiskFileReader. Now you can pass in a quarantine_hook callable to DiskFile.reader() that gets called if the file was quarantined. Default is to log quarantines, but otherwise do nothing. Change-Id: I04fc14569982a17fcc89e00832725ae71009335a
2013-10-28 14:57:18 -07:00
quarantine_msgs = []
reader = df.reader(_quarantine_hook=quarantine_msgs.append)
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
self.assertEquals(''.join(reader.app_iter_range(0, None)),
'1234567890')
Alternate DiskFile constructor for efficient auditing. Before, to audit an object, the auditor: - calls listdir(object-hash-dir) - picks out the .data file from the listing - pulls out all N of its user.swift.metadata* xattrs - unpickles them - pulls out the value for 'name' - splits the name into a/c/o - then instantiates and opens a DiskFile(a, c, o), which does the following - joins a/c/o back into a name - hashes the name - calls listdir(object-hash-dir) (AGAIN) - picks out the .data file (and maybe .meta) from the listing (AGAIN) - pulls out all N of its user.swift.metadata* xattrs (AGAIN) - unpickles them (AGAIN) - starts reading object's contents off disk Now, the auditor simply locates the hash dir on the filesystem (saving one listdir) and then hands it off to DiskFileManager.get_diskfile_from_audit_location, which then instantiates a DiskFile in a way that lazy-loads the name later (saving one xattr reading). As part of this, DiskFile.open() will now quarantine a hash "directory" that's actually a file. Before, the audit location generator would skip those, but now they make it clear into DiskFile(). It's better to quarantine them anyway, as they're not doing any good the way they are. Also, removed the was_quarantined attribute on DiskFileReader. Now you can pass in a quarantine_hook callable to DiskFile.reader() that gets called if the file was quarantined. Default is to log quarantines, but otherwise do nothing. Change-Id: I04fc14569982a17fcc89e00832725ae71009335a
2013-10-28 14:57:18 -07:00
self.assertEquals(quarantine_msgs, [])
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda', '0', 'a', 'c', 'o')
with df.open():
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
reader = df.reader()
self.assertEqual(''.join(reader.app_iter_range(5, None)), '67890')
def test_disk_file_app_iter_partial_closes(self):
df = self._create_test_file('1234567890')
Alternate DiskFile constructor for efficient auditing. Before, to audit an object, the auditor: - calls listdir(object-hash-dir) - picks out the .data file from the listing - pulls out all N of its user.swift.metadata* xattrs - unpickles them - pulls out the value for 'name' - splits the name into a/c/o - then instantiates and opens a DiskFile(a, c, o), which does the following - joins a/c/o back into a name - hashes the name - calls listdir(object-hash-dir) (AGAIN) - picks out the .data file (and maybe .meta) from the listing (AGAIN) - pulls out all N of its user.swift.metadata* xattrs (AGAIN) - unpickles them (AGAIN) - starts reading object's contents off disk Now, the auditor simply locates the hash dir on the filesystem (saving one listdir) and then hands it off to DiskFileManager.get_diskfile_from_audit_location, which then instantiates a DiskFile in a way that lazy-loads the name later (saving one xattr reading). As part of this, DiskFile.open() will now quarantine a hash "directory" that's actually a file. Before, the audit location generator would skip those, but now they make it clear into DiskFile(). It's better to quarantine them anyway, as they're not doing any good the way they are. Also, removed the was_quarantined attribute on DiskFileReader. Now you can pass in a quarantine_hook callable to DiskFile.reader() that gets called if the file was quarantined. Default is to log quarantines, but otherwise do nothing. Change-Id: I04fc14569982a17fcc89e00832725ae71009335a
2013-10-28 14:57:18 -07:00
quarantine_msgs = []
reader = df.reader(_quarantine_hook=quarantine_msgs.append)
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
it = reader.app_iter_range(0, 5)
Alternate DiskFile constructor for efficient auditing. Before, to audit an object, the auditor: - calls listdir(object-hash-dir) - picks out the .data file from the listing - pulls out all N of its user.swift.metadata* xattrs - unpickles them - pulls out the value for 'name' - splits the name into a/c/o - then instantiates and opens a DiskFile(a, c, o), which does the following - joins a/c/o back into a name - hashes the name - calls listdir(object-hash-dir) (AGAIN) - picks out the .data file (and maybe .meta) from the listing (AGAIN) - pulls out all N of its user.swift.metadata* xattrs (AGAIN) - unpickles them (AGAIN) - starts reading object's contents off disk Now, the auditor simply locates the hash dir on the filesystem (saving one listdir) and then hands it off to DiskFileManager.get_diskfile_from_audit_location, which then instantiates a DiskFile in a way that lazy-loads the name later (saving one xattr reading). As part of this, DiskFile.open() will now quarantine a hash "directory" that's actually a file. Before, the audit location generator would skip those, but now they make it clear into DiskFile(). It's better to quarantine them anyway, as they're not doing any good the way they are. Also, removed the was_quarantined attribute on DiskFileReader. Now you can pass in a quarantine_hook callable to DiskFile.reader() that gets called if the file was quarantined. Default is to log quarantines, but otherwise do nothing. Change-Id: I04fc14569982a17fcc89e00832725ae71009335a
2013-10-28 14:57:18 -07:00
self.assertEquals(quarantine_msgs, [])
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
self.assertEqual(''.join(it), '12345')
self.assertTrue(reader._fp is None)
def test_disk_file_app_iter_ranges(self):
df = self._create_test_file('012345678911234567892123456789')
Alternate DiskFile constructor for efficient auditing. Before, to audit an object, the auditor: - calls listdir(object-hash-dir) - picks out the .data file from the listing - pulls out all N of its user.swift.metadata* xattrs - unpickles them - pulls out the value for 'name' - splits the name into a/c/o - then instantiates and opens a DiskFile(a, c, o), which does the following - joins a/c/o back into a name - hashes the name - calls listdir(object-hash-dir) (AGAIN) - picks out the .data file (and maybe .meta) from the listing (AGAIN) - pulls out all N of its user.swift.metadata* xattrs (AGAIN) - unpickles them (AGAIN) - starts reading object's contents off disk Now, the auditor simply locates the hash dir on the filesystem (saving one listdir) and then hands it off to DiskFileManager.get_diskfile_from_audit_location, which then instantiates a DiskFile in a way that lazy-loads the name later (saving one xattr reading). As part of this, DiskFile.open() will now quarantine a hash "directory" that's actually a file. Before, the audit location generator would skip those, but now they make it clear into DiskFile(). It's better to quarantine them anyway, as they're not doing any good the way they are. Also, removed the was_quarantined attribute on DiskFileReader. Now you can pass in a quarantine_hook callable to DiskFile.reader() that gets called if the file was quarantined. Default is to log quarantines, but otherwise do nothing. Change-Id: I04fc14569982a17fcc89e00832725ae71009335a
2013-10-28 14:57:18 -07:00
quarantine_msgs = []
reader = df.reader(_quarantine_hook=quarantine_msgs.append)
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
it = reader.app_iter_ranges([(0, 10), (10, 20), (20, 30)],
'plain/text',
'\r\n--someheader\r\n', 30)
Alternate DiskFile constructor for efficient auditing. Before, to audit an object, the auditor: - calls listdir(object-hash-dir) - picks out the .data file from the listing - pulls out all N of its user.swift.metadata* xattrs - unpickles them - pulls out the value for 'name' - splits the name into a/c/o - then instantiates and opens a DiskFile(a, c, o), which does the following - joins a/c/o back into a name - hashes the name - calls listdir(object-hash-dir) (AGAIN) - picks out the .data file (and maybe .meta) from the listing (AGAIN) - pulls out all N of its user.swift.metadata* xattrs (AGAIN) - unpickles them (AGAIN) - starts reading object's contents off disk Now, the auditor simply locates the hash dir on the filesystem (saving one listdir) and then hands it off to DiskFileManager.get_diskfile_from_audit_location, which then instantiates a DiskFile in a way that lazy-loads the name later (saving one xattr reading). As part of this, DiskFile.open() will now quarantine a hash "directory" that's actually a file. Before, the audit location generator would skip those, but now they make it clear into DiskFile(). It's better to quarantine them anyway, as they're not doing any good the way they are. Also, removed the was_quarantined attribute on DiskFileReader. Now you can pass in a quarantine_hook callable to DiskFile.reader() that gets called if the file was quarantined. Default is to log quarantines, but otherwise do nothing. Change-Id: I04fc14569982a17fcc89e00832725ae71009335a
2013-10-28 14:57:18 -07:00
self.assertEquals(quarantine_msgs, [])
value = ''.join(it)
self.assert_('0123456789' in value)
self.assert_('1123456789' in value)
self.assert_('2123456789' in value)
def test_disk_file_app_iter_ranges_edges(self):
df = self._create_test_file('012345678911234567892123456789')
Alternate DiskFile constructor for efficient auditing. Before, to audit an object, the auditor: - calls listdir(object-hash-dir) - picks out the .data file from the listing - pulls out all N of its user.swift.metadata* xattrs - unpickles them - pulls out the value for 'name' - splits the name into a/c/o - then instantiates and opens a DiskFile(a, c, o), which does the following - joins a/c/o back into a name - hashes the name - calls listdir(object-hash-dir) (AGAIN) - picks out the .data file (and maybe .meta) from the listing (AGAIN) - pulls out all N of its user.swift.metadata* xattrs (AGAIN) - unpickles them (AGAIN) - starts reading object's contents off disk Now, the auditor simply locates the hash dir on the filesystem (saving one listdir) and then hands it off to DiskFileManager.get_diskfile_from_audit_location, which then instantiates a DiskFile in a way that lazy-loads the name later (saving one xattr reading). As part of this, DiskFile.open() will now quarantine a hash "directory" that's actually a file. Before, the audit location generator would skip those, but now they make it clear into DiskFile(). It's better to quarantine them anyway, as they're not doing any good the way they are. Also, removed the was_quarantined attribute on DiskFileReader. Now you can pass in a quarantine_hook callable to DiskFile.reader() that gets called if the file was quarantined. Default is to log quarantines, but otherwise do nothing. Change-Id: I04fc14569982a17fcc89e00832725ae71009335a
2013-10-28 14:57:18 -07:00
quarantine_msgs = []
reader = df.reader(_quarantine_hook=quarantine_msgs.append)
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
it = reader.app_iter_ranges([(3, 10), (0, 2)], 'application/whatever',
'\r\n--someheader\r\n', 30)
value = ''.join(it)
Alternate DiskFile constructor for efficient auditing. Before, to audit an object, the auditor: - calls listdir(object-hash-dir) - picks out the .data file from the listing - pulls out all N of its user.swift.metadata* xattrs - unpickles them - pulls out the value for 'name' - splits the name into a/c/o - then instantiates and opens a DiskFile(a, c, o), which does the following - joins a/c/o back into a name - hashes the name - calls listdir(object-hash-dir) (AGAIN) - picks out the .data file (and maybe .meta) from the listing (AGAIN) - pulls out all N of its user.swift.metadata* xattrs (AGAIN) - unpickles them (AGAIN) - starts reading object's contents off disk Now, the auditor simply locates the hash dir on the filesystem (saving one listdir) and then hands it off to DiskFileManager.get_diskfile_from_audit_location, which then instantiates a DiskFile in a way that lazy-loads the name later (saving one xattr reading). As part of this, DiskFile.open() will now quarantine a hash "directory" that's actually a file. Before, the audit location generator would skip those, but now they make it clear into DiskFile(). It's better to quarantine them anyway, as they're not doing any good the way they are. Also, removed the was_quarantined attribute on DiskFileReader. Now you can pass in a quarantine_hook callable to DiskFile.reader() that gets called if the file was quarantined. Default is to log quarantines, but otherwise do nothing. Change-Id: I04fc14569982a17fcc89e00832725ae71009335a
2013-10-28 14:57:18 -07:00
self.assertEquals(quarantine_msgs, [])
self.assert_('3456789' in value)
self.assert_('01' in value)
def test_disk_file_large_app_iter_ranges(self):
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
# This test case is to make sure that the disk file app_iter_ranges
# method all the paths being tested.
long_str = '01234567890' * 65536
target_strs = ['3456789', long_str[0:65590]]
df = self._create_test_file(long_str)
Alternate DiskFile constructor for efficient auditing. Before, to audit an object, the auditor: - calls listdir(object-hash-dir) - picks out the .data file from the listing - pulls out all N of its user.swift.metadata* xattrs - unpickles them - pulls out the value for 'name' - splits the name into a/c/o - then instantiates and opens a DiskFile(a, c, o), which does the following - joins a/c/o back into a name - hashes the name - calls listdir(object-hash-dir) (AGAIN) - picks out the .data file (and maybe .meta) from the listing (AGAIN) - pulls out all N of its user.swift.metadata* xattrs (AGAIN) - unpickles them (AGAIN) - starts reading object's contents off disk Now, the auditor simply locates the hash dir on the filesystem (saving one listdir) and then hands it off to DiskFileManager.get_diskfile_from_audit_location, which then instantiates a DiskFile in a way that lazy-loads the name later (saving one xattr reading). As part of this, DiskFile.open() will now quarantine a hash "directory" that's actually a file. Before, the audit location generator would skip those, but now they make it clear into DiskFile(). It's better to quarantine them anyway, as they're not doing any good the way they are. Also, removed the was_quarantined attribute on DiskFileReader. Now you can pass in a quarantine_hook callable to DiskFile.reader() that gets called if the file was quarantined. Default is to log quarantines, but otherwise do nothing. Change-Id: I04fc14569982a17fcc89e00832725ae71009335a
2013-10-28 14:57:18 -07:00
quarantine_msgs = []
reader = df.reader(_quarantine_hook=quarantine_msgs.append)
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
it = reader.app_iter_ranges([(3, 10), (0, 65590)], 'plain/text',
'5e816ff8b8b8e9a5d355497e5d9e0301', 655360)
# The produced string actually missing the MIME headers
# need to add these headers to make it as real MIME message.
# The body of the message is produced by method app_iter_ranges
# off of DiskFile object.
header = ''.join(['Content-Type: multipart/byteranges;',
'boundary=',
'5e816ff8b8b8e9a5d355497e5d9e0301\r\n'])
value = header + ''.join(it)
Alternate DiskFile constructor for efficient auditing. Before, to audit an object, the auditor: - calls listdir(object-hash-dir) - picks out the .data file from the listing - pulls out all N of its user.swift.metadata* xattrs - unpickles them - pulls out the value for 'name' - splits the name into a/c/o - then instantiates and opens a DiskFile(a, c, o), which does the following - joins a/c/o back into a name - hashes the name - calls listdir(object-hash-dir) (AGAIN) - picks out the .data file (and maybe .meta) from the listing (AGAIN) - pulls out all N of its user.swift.metadata* xattrs (AGAIN) - unpickles them (AGAIN) - starts reading object's contents off disk Now, the auditor simply locates the hash dir on the filesystem (saving one listdir) and then hands it off to DiskFileManager.get_diskfile_from_audit_location, which then instantiates a DiskFile in a way that lazy-loads the name later (saving one xattr reading). As part of this, DiskFile.open() will now quarantine a hash "directory" that's actually a file. Before, the audit location generator would skip those, but now they make it clear into DiskFile(). It's better to quarantine them anyway, as they're not doing any good the way they are. Also, removed the was_quarantined attribute on DiskFileReader. Now you can pass in a quarantine_hook callable to DiskFile.reader() that gets called if the file was quarantined. Default is to log quarantines, but otherwise do nothing. Change-Id: I04fc14569982a17fcc89e00832725ae71009335a
2013-10-28 14:57:18 -07:00
self.assertEquals(quarantine_msgs, [])
parts = map(lambda p: p.get_payload(decode=True),
email.message_from_string(value).walk())[1:3]
self.assertEqual(parts, target_strs)
def test_disk_file_app_iter_ranges_empty(self):
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
# This test case tests when empty value passed into app_iter_ranges
# When ranges passed into the method is either empty array or None,
# this method will yield empty string
df = self._create_test_file('012345678911234567892123456789')
Alternate DiskFile constructor for efficient auditing. Before, to audit an object, the auditor: - calls listdir(object-hash-dir) - picks out the .data file from the listing - pulls out all N of its user.swift.metadata* xattrs - unpickles them - pulls out the value for 'name' - splits the name into a/c/o - then instantiates and opens a DiskFile(a, c, o), which does the following - joins a/c/o back into a name - hashes the name - calls listdir(object-hash-dir) (AGAIN) - picks out the .data file (and maybe .meta) from the listing (AGAIN) - pulls out all N of its user.swift.metadata* xattrs (AGAIN) - unpickles them (AGAIN) - starts reading object's contents off disk Now, the auditor simply locates the hash dir on the filesystem (saving one listdir) and then hands it off to DiskFileManager.get_diskfile_from_audit_location, which then instantiates a DiskFile in a way that lazy-loads the name later (saving one xattr reading). As part of this, DiskFile.open() will now quarantine a hash "directory" that's actually a file. Before, the audit location generator would skip those, but now they make it clear into DiskFile(). It's better to quarantine them anyway, as they're not doing any good the way they are. Also, removed the was_quarantined attribute on DiskFileReader. Now you can pass in a quarantine_hook callable to DiskFile.reader() that gets called if the file was quarantined. Default is to log quarantines, but otherwise do nothing. Change-Id: I04fc14569982a17fcc89e00832725ae71009335a
2013-10-28 14:57:18 -07:00
quarantine_msgs = []
reader = df.reader(_quarantine_hook=quarantine_msgs.append)
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
it = reader.app_iter_ranges([], 'application/whatever',
'\r\n--someheader\r\n', 100)
self.assertEqual(''.join(it), '')
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda', '0', 'a', 'c', 'o')
with df.open():
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
reader = df.reader()
it = reader.app_iter_ranges(None, 'app/something',
'\r\n--someheader\r\n', 150)
Alternate DiskFile constructor for efficient auditing. Before, to audit an object, the auditor: - calls listdir(object-hash-dir) - picks out the .data file from the listing - pulls out all N of its user.swift.metadata* xattrs - unpickles them - pulls out the value for 'name' - splits the name into a/c/o - then instantiates and opens a DiskFile(a, c, o), which does the following - joins a/c/o back into a name - hashes the name - calls listdir(object-hash-dir) (AGAIN) - picks out the .data file (and maybe .meta) from the listing (AGAIN) - pulls out all N of its user.swift.metadata* xattrs (AGAIN) - unpickles them (AGAIN) - starts reading object's contents off disk Now, the auditor simply locates the hash dir on the filesystem (saving one listdir) and then hands it off to DiskFileManager.get_diskfile_from_audit_location, which then instantiates a DiskFile in a way that lazy-loads the name later (saving one xattr reading). As part of this, DiskFile.open() will now quarantine a hash "directory" that's actually a file. Before, the audit location generator would skip those, but now they make it clear into DiskFile(). It's better to quarantine them anyway, as they're not doing any good the way they are. Also, removed the was_quarantined attribute on DiskFileReader. Now you can pass in a quarantine_hook callable to DiskFile.reader() that gets called if the file was quarantined. Default is to log quarantines, but otherwise do nothing. Change-Id: I04fc14569982a17fcc89e00832725ae71009335a
2013-10-28 14:57:18 -07:00
self.assertEquals(quarantine_msgs, [])
self.assertEqual(''.join(it), '')
def test_disk_file_mkstemp_creates_dir(self):
tmpdir = os.path.join(self.testdir, 'sda1', 'tmp')
os.rmdir(tmpdir)
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
with df.create():
self.assert_(os.path.exists(tmpdir))
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
def _get_open_disk_file(self, invalid_type=None, obj_name='o', fsize=1024,
csize=8, mark_deleted=False, ts=None,
mount_check=False, extra_metadata=None):
'''returns a DiskFile'''
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', obj_name)
data = '0' * fsize
etag = md5()
if ts:
timestamp = ts
else:
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
timestamp = normalize_timestamp(time())
with df.create() as writer:
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
upload_size = writer.write(data)
etag.update(data)
etag = etag.hexdigest()
metadata = {
'ETag': etag,
'X-Timestamp': timestamp,
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
'Content-Length': str(upload_size),
}
metadata.update(extra_metadata or {})
writer.put(metadata)
if invalid_type == 'ETag':
etag = md5()
etag.update('1' + '0' * (fsize - 1))
etag = etag.hexdigest()
metadata['ETag'] = etag
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
diskfile.write_metadata(writer._fd, metadata)
elif invalid_type == 'Content-Length':
metadata['Content-Length'] = fsize - 1
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
diskfile.write_metadata(writer._fd, metadata)
elif invalid_type == 'Bad-Content-Length':
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
metadata['Content-Length'] = 'zero'
diskfile.write_metadata(writer._fd, metadata)
elif invalid_type == 'Missing-Content-Length':
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
del metadata['Content-Length']
diskfile.write_metadata(writer._fd, metadata)
if mark_deleted:
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df.delete(timestamp)
data_files = [os.path.join(df._datadir, fname)
for fname in sorted(os.listdir(df._datadir),
reverse=True)
if fname.endswith('.data')]
if invalid_type == 'Corrupt-Xattrs':
# We have to go below read_metadata/write_metadata to get proper
# corruption.
meta_xattr = xattr.getxattr(data_files[0], "user.swift.metadata")
wrong_byte = 'X' if meta_xattr[0] != 'X' else 'Y'
xattr.setxattr(data_files[0], "user.swift.metadata",
wrong_byte + meta_xattr[1:])
elif invalid_type == 'Truncated-Xattrs':
meta_xattr = xattr.getxattr(data_files[0], "user.swift.metadata")
xattr.setxattr(data_files[0], "user.swift.metadata",
meta_xattr[:-1])
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
self.conf['disk_chunk_size'] = csize
self.conf['mount_check'] = mount_check
self.df_mgr = diskfile.DiskFileManager(self.conf, FakeLogger())
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', obj_name)
df.open()
if invalid_type == 'Zero-Byte':
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
fp = open(df._data_file, 'w')
fp.close()
df.unit_test_len = fsize
return df
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
def test_keep_cache(self):
df = self._get_open_disk_file(fsize=65)
with mock.patch("swift.obj.diskfile.drop_buffer_cache") as foo:
for _ in df.reader():
pass
self.assertTrue(foo.called)
df = self._get_open_disk_file(fsize=65)
with mock.patch("swift.obj.diskfile.drop_buffer_cache") as bar:
for _ in df.reader(keep_cache=False):
pass
self.assertTrue(bar.called)
df = self._get_open_disk_file(fsize=65)
with mock.patch("swift.obj.diskfile.drop_buffer_cache") as boo:
for _ in df.reader(keep_cache=True):
pass
self.assertFalse(boo.called)
df = self._get_open_disk_file(fsize=5 * 1024, csize=256)
with mock.patch("swift.obj.diskfile.drop_buffer_cache") as goo:
for _ in df.reader(keep_cache=True):
pass
self.assertTrue(goo.called)
def test_quarantine_valids(self):
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
def verify(*args, **kwargs):
try:
df = self._get_open_disk_file(**kwargs)
reader = df.reader()
for chunk in reader:
pass
except DiskFileQuarantined:
self.fail(
"Unexpected quarantining occurred: args=%r, kwargs=%r" % (
args, kwargs))
else:
pass
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
verify(obj_name='1')
verify(obj_name='2', csize=1)
verify(obj_name='3', csize=100000)
def run_quarantine_invalids(self, invalid_type):
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
def verify(*args, **kwargs):
open_exc = invalid_type in ('Content-Length', 'Bad-Content-Length',
'Corrupt-Xattrs', 'Truncated-Xattrs')
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
reader = None
Alternate DiskFile constructor for efficient auditing. Before, to audit an object, the auditor: - calls listdir(object-hash-dir) - picks out the .data file from the listing - pulls out all N of its user.swift.metadata* xattrs - unpickles them - pulls out the value for 'name' - splits the name into a/c/o - then instantiates and opens a DiskFile(a, c, o), which does the following - joins a/c/o back into a name - hashes the name - calls listdir(object-hash-dir) (AGAIN) - picks out the .data file (and maybe .meta) from the listing (AGAIN) - pulls out all N of its user.swift.metadata* xattrs (AGAIN) - unpickles them (AGAIN) - starts reading object's contents off disk Now, the auditor simply locates the hash dir on the filesystem (saving one listdir) and then hands it off to DiskFileManager.get_diskfile_from_audit_location, which then instantiates a DiskFile in a way that lazy-loads the name later (saving one xattr reading). As part of this, DiskFile.open() will now quarantine a hash "directory" that's actually a file. Before, the audit location generator would skip those, but now they make it clear into DiskFile(). It's better to quarantine them anyway, as they're not doing any good the way they are. Also, removed the was_quarantined attribute on DiskFileReader. Now you can pass in a quarantine_hook callable to DiskFile.reader() that gets called if the file was quarantined. Default is to log quarantines, but otherwise do nothing. Change-Id: I04fc14569982a17fcc89e00832725ae71009335a
2013-10-28 14:57:18 -07:00
quarantine_msgs = []
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
try:
df = self._get_open_disk_file(**kwargs)
Alternate DiskFile constructor for efficient auditing. Before, to audit an object, the auditor: - calls listdir(object-hash-dir) - picks out the .data file from the listing - pulls out all N of its user.swift.metadata* xattrs - unpickles them - pulls out the value for 'name' - splits the name into a/c/o - then instantiates and opens a DiskFile(a, c, o), which does the following - joins a/c/o back into a name - hashes the name - calls listdir(object-hash-dir) (AGAIN) - picks out the .data file (and maybe .meta) from the listing (AGAIN) - pulls out all N of its user.swift.metadata* xattrs (AGAIN) - unpickles them (AGAIN) - starts reading object's contents off disk Now, the auditor simply locates the hash dir on the filesystem (saving one listdir) and then hands it off to DiskFileManager.get_diskfile_from_audit_location, which then instantiates a DiskFile in a way that lazy-loads the name later (saving one xattr reading). As part of this, DiskFile.open() will now quarantine a hash "directory" that's actually a file. Before, the audit location generator would skip those, but now they make it clear into DiskFile(). It's better to quarantine them anyway, as they're not doing any good the way they are. Also, removed the was_quarantined attribute on DiskFileReader. Now you can pass in a quarantine_hook callable to DiskFile.reader() that gets called if the file was quarantined. Default is to log quarantines, but otherwise do nothing. Change-Id: I04fc14569982a17fcc89e00832725ae71009335a
2013-10-28 14:57:18 -07:00
reader = df.reader(_quarantine_hook=quarantine_msgs.append)
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
except DiskFileQuarantined as err:
if not open_exc:
self.fail(
"Unexpected DiskFileQuarantine raised: :%r" % err)
return
else:
if open_exc:
self.fail("Expected DiskFileQuarantine exception")
try:
for chunk in reader:
pass
except DiskFileQuarantined as err:
self.fail("Unexpected DiskFileQuarantine raised: :%r" % err)
else:
if not open_exc:
Alternate DiskFile constructor for efficient auditing. Before, to audit an object, the auditor: - calls listdir(object-hash-dir) - picks out the .data file from the listing - pulls out all N of its user.swift.metadata* xattrs - unpickles them - pulls out the value for 'name' - splits the name into a/c/o - then instantiates and opens a DiskFile(a, c, o), which does the following - joins a/c/o back into a name - hashes the name - calls listdir(object-hash-dir) (AGAIN) - picks out the .data file (and maybe .meta) from the listing (AGAIN) - pulls out all N of its user.swift.metadata* xattrs (AGAIN) - unpickles them (AGAIN) - starts reading object's contents off disk Now, the auditor simply locates the hash dir on the filesystem (saving one listdir) and then hands it off to DiskFileManager.get_diskfile_from_audit_location, which then instantiates a DiskFile in a way that lazy-loads the name later (saving one xattr reading). As part of this, DiskFile.open() will now quarantine a hash "directory" that's actually a file. Before, the audit location generator would skip those, but now they make it clear into DiskFile(). It's better to quarantine them anyway, as they're not doing any good the way they are. Also, removed the was_quarantined attribute on DiskFileReader. Now you can pass in a quarantine_hook callable to DiskFile.reader() that gets called if the file was quarantined. Default is to log quarantines, but otherwise do nothing. Change-Id: I04fc14569982a17fcc89e00832725ae71009335a
2013-10-28 14:57:18 -07:00
self.assertEqual(1, len(quarantine_msgs))
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
verify(invalid_type=invalid_type, obj_name='1')
verify(invalid_type=invalid_type, obj_name='2', csize=1)
verify(invalid_type=invalid_type, obj_name='3', csize=100000)
verify(invalid_type=invalid_type, obj_name='4')
def verify_air(params, start=0, adjustment=0):
"""verify (a)pp (i)ter (r)ange"""
open_exc = invalid_type in ('Content-Length', 'Bad-Content-Length',
'Corrupt-Xattrs', 'Truncated-Xattrs')
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
reader = None
try:
df = self._get_open_disk_file(**params)
reader = df.reader()
except DiskFileQuarantined as err:
if not open_exc:
self.fail(
"Unexpected DiskFileQuarantine raised: :%r" % err)
return
else:
if open_exc:
self.fail("Expected DiskFileQuarantine exception")
try:
for chunk in reader.app_iter_range(
start,
df.unit_test_len + adjustment):
pass
except DiskFileQuarantined as err:
self.fail("Unexpected DiskFileQuarantine raised: :%r" % err)
verify_air(dict(invalid_type=invalid_type, obj_name='5'))
verify_air(dict(invalid_type=invalid_type, obj_name='6'), 0, 100)
verify_air(dict(invalid_type=invalid_type, obj_name='7'), 1)
verify_air(dict(invalid_type=invalid_type, obj_name='8'), 0, -1)
verify_air(dict(invalid_type=invalid_type, obj_name='8'), 1, 1)
def test_quarantine_corrupt_xattrs(self):
self.run_quarantine_invalids('Corrupt-Xattrs')
def test_quarantine_truncated_xattrs(self):
self.run_quarantine_invalids('Truncated-Xattrs')
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
def test_quarantine_invalid_etag(self):
self.run_quarantine_invalids('ETag')
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
def test_quarantine_invalid_content_length(self):
self.run_quarantine_invalids('Content-Length')
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
def test_quarantine_invalid_content_length_bad(self):
self.run_quarantine_invalids('Bad-Content-Length')
def test_quarantine_invalid_zero_byte(self):
self.run_quarantine_invalids('Zero-Byte')
def test_quarantine_deleted_files(self):
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
try:
self._get_open_disk_file(invalid_type='Content-Length')
except DiskFileQuarantined:
pass
else:
self.fail("Expected DiskFileQuarantined exception")
try:
self._get_open_disk_file(invalid_type='Content-Length',
mark_deleted=True)
except DiskFileQuarantined as err:
self.fail("Unexpected DiskFileQuarantined exception"
" encountered: %r" % err)
except DiskFileNotExist:
pass
else:
self.fail("Expected DiskFileNotExist exception")
try:
self._get_open_disk_file(invalid_type='Content-Length',
mark_deleted=True)
except DiskFileNotExist:
pass
else:
self.fail("Expected DiskFileNotExist exception")
def test_quarantine_missing_content_length(self):
try:
self._get_open_disk_file(
invalid_type='Missing-Content-Length')
except DiskFileQuarantined:
pass
def test_quarantine_bad_content_length(self):
try:
self._get_open_disk_file(
invalid_type='Bad-Content-Length')
except DiskFileQuarantined:
pass
else:
self.fail("Expected DiskFileQuarantined exception")
def test_quarantine_fstat_oserror(self):
invocations = [0]
orig_os_fstat = os.fstat
def bad_fstat(fd):
invocations[0] += 1
if invocations[0] == 4:
# FIXME - yes, this an icky way to get code coverage ... worth
# it?
raise OSError()
return orig_os_fstat(fd)
with mock.patch('os.fstat', bad_fstat):
try:
self._get_open_disk_file()
except DiskFileQuarantined:
pass
else:
self.fail("Expected DiskFileQuarantined exception")
Alternate DiskFile constructor for efficient auditing. Before, to audit an object, the auditor: - calls listdir(object-hash-dir) - picks out the .data file from the listing - pulls out all N of its user.swift.metadata* xattrs - unpickles them - pulls out the value for 'name' - splits the name into a/c/o - then instantiates and opens a DiskFile(a, c, o), which does the following - joins a/c/o back into a name - hashes the name - calls listdir(object-hash-dir) (AGAIN) - picks out the .data file (and maybe .meta) from the listing (AGAIN) - pulls out all N of its user.swift.metadata* xattrs (AGAIN) - unpickles them (AGAIN) - starts reading object's contents off disk Now, the auditor simply locates the hash dir on the filesystem (saving one listdir) and then hands it off to DiskFileManager.get_diskfile_from_audit_location, which then instantiates a DiskFile in a way that lazy-loads the name later (saving one xattr reading). As part of this, DiskFile.open() will now quarantine a hash "directory" that's actually a file. Before, the audit location generator would skip those, but now they make it clear into DiskFile(). It's better to quarantine them anyway, as they're not doing any good the way they are. Also, removed the was_quarantined attribute on DiskFileReader. Now you can pass in a quarantine_hook callable to DiskFile.reader() that gets called if the file was quarantined. Default is to log quarantines, but otherwise do nothing. Change-Id: I04fc14569982a17fcc89e00832725ae71009335a
2013-10-28 14:57:18 -07:00
def test_quarantine_hashdir_not_a_directory(self):
df = self._create_test_file('1234567890', account="abc",
container='123', object='xyz')
hashdir = df._datadir
rmtree(hashdir)
with open(hashdir, 'w'):
pass
df = self.df_mgr.get_diskfile('sda', '0', 'abc', '123', 'xyz')
try:
df.open()
except DiskFileQuarantined:
pass
else:
self.fail("Expected DiskFileQuarantined, didn't get it")
# make sure the right thing got quarantined; the suffix dir should not
# have moved, as that could have many objects in it
self.assertFalse(os.path.exists(hashdir))
self.assertTrue(os.path.exists(os.path.dirname(hashdir)))
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
def test_write_metadata(self):
df = self._create_test_file('1234567890')
timestamp = normalize_timestamp(time())
metadata = {'X-Timestamp': timestamp, 'X-Object-Meta-test': 'data'}
df.write_metadata(metadata)
dl = os.listdir(df._datadir)
self.assertEquals(len(dl), 2)
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
exp_name = '%s.meta' % timestamp
self.assertTrue(exp_name in set(dl))
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
def test_delete(self):
df = self._get_open_disk_file()
ts = time()
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df.delete(ts)
exp_name = '%s.ts' % str(normalize_timestamp(ts))
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
dl = os.listdir(df._datadir)
self.assertEquals(len(dl), 1)
self.assertTrue(exp_name in set(dl))
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
def test_open_deleted(self):
df = self._get_open_disk_file()
ts = time()
df.delete(ts)
exp_name = '%s.ts' % str(normalize_timestamp(ts))
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
dl = os.listdir(df._datadir)
self.assertEquals(len(dl), 1)
self.assertTrue(exp_name in set(dl))
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
self.assertRaises(DiskFileDeleted, df.open)
def test_open_deleted_with_corrupt_tombstone(self):
df = self._get_open_disk_file()
ts = time()
df.delete(ts)
exp_name = '%s.ts' % str(normalize_timestamp(ts))
dl = os.listdir(df._datadir)
self.assertEquals(len(dl), 1)
self.assertTrue(exp_name in set(dl))
# it's pickle-format, so removing the last byte is sufficient to
# corrupt it
ts_fullpath = os.path.join(df._datadir, exp_name)
self.assertTrue(os.path.exists(ts_fullpath)) # sanity check
meta_xattr = xattr.getxattr(ts_fullpath, "user.swift.metadata")
xattr.setxattr(ts_fullpath, "user.swift.metadata", meta_xattr[:-1])
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
self.assertRaises(DiskFileNotExist, df.open)
self.assertFalse(os.path.exists(ts_fullpath))
Alternate DiskFile constructor for efficient auditing. Before, to audit an object, the auditor: - calls listdir(object-hash-dir) - picks out the .data file from the listing - pulls out all N of its user.swift.metadata* xattrs - unpickles them - pulls out the value for 'name' - splits the name into a/c/o - then instantiates and opens a DiskFile(a, c, o), which does the following - joins a/c/o back into a name - hashes the name - calls listdir(object-hash-dir) (AGAIN) - picks out the .data file (and maybe .meta) from the listing (AGAIN) - pulls out all N of its user.swift.metadata* xattrs (AGAIN) - unpickles them (AGAIN) - starts reading object's contents off disk Now, the auditor simply locates the hash dir on the filesystem (saving one listdir) and then hands it off to DiskFileManager.get_diskfile_from_audit_location, which then instantiates a DiskFile in a way that lazy-loads the name later (saving one xattr reading). As part of this, DiskFile.open() will now quarantine a hash "directory" that's actually a file. Before, the audit location generator would skip those, but now they make it clear into DiskFile(). It's better to quarantine them anyway, as they're not doing any good the way they are. Also, removed the was_quarantined attribute on DiskFileReader. Now you can pass in a quarantine_hook callable to DiskFile.reader() that gets called if the file was quarantined. Default is to log quarantines, but otherwise do nothing. Change-Id: I04fc14569982a17fcc89e00832725ae71009335a
2013-10-28 14:57:18 -07:00
def test_from_audit_location(self):
hashdir = self._create_test_file(
'blah blah',
account='three', container='blind', object='mice')._datadir
df = self.df_mgr.get_diskfile_from_audit_location(
diskfile.AuditLocation(hashdir, 'sda1', '0'))
df.open()
self.assertEqual(df._name, '/three/blind/mice')
def test_from_audit_location_with_mismatched_hash(self):
hashdir = self._create_test_file(
'blah blah',
account='this', container='is', object='right')._datadir
datafile = os.path.join(hashdir, os.listdir(hashdir)[0])
meta = diskfile.read_metadata(datafile)
meta['name'] = '/this/is/wrong'
diskfile.write_metadata(datafile, meta)
df = self.df_mgr.get_diskfile_from_audit_location(
diskfile.AuditLocation(hashdir, 'sda1', '0'))
self.assertRaises(DiskFileQuarantined, df.open)
def test_close_error(self):
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
def mock_handle_close_quarantine():
raise Exception("Bad")
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self._get_open_disk_file(fsize=1024 * 1024 * 2)
reader = df.reader()
reader._handle_close_quarantine = mock_handle_close_quarantine
for chunk in reader:
pass
# close is called at the end of the iterator
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
self.assertEquals(reader._fp, None)
self.assertEquals(len(df._logger.log_dict['error']), 1)
def test_mount_checking(self):
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
def _mock_cm(*args, **kwargs):
return False
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
with mock.patch("swift.common.constraints.check_mount", _mock_cm):
try:
self._get_open_disk_file(mount_check=True)
except DiskFileDeviceUnavailable:
pass
else:
self.fail("Expected DiskFileDeviceUnavailable exception")
def test_ondisk_search_loop_ts_meta_data(self):
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
self._create_ondisk_file(df, '', ext='.ts', timestamp=10)
self._create_ondisk_file(df, '', ext='.ts', timestamp=9)
self._create_ondisk_file(df, '', ext='.meta', timestamp=8)
self._create_ondisk_file(df, '', ext='.meta', timestamp=7)
self._create_ondisk_file(df, 'B', ext='.data', timestamp=6)
self._create_ondisk_file(df, 'A', ext='.data', timestamp=5)
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
try:
df.open()
except DiskFileDeleted as d:
self.assertEquals(d.timestamp, normalize_timestamp(10))
else:
self.fail("Expected DiskFileDeleted exception")
def test_ondisk_search_loop_meta_ts_data(self):
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
self._create_ondisk_file(df, '', ext='.meta', timestamp=10)
self._create_ondisk_file(df, '', ext='.meta', timestamp=9)
self._create_ondisk_file(df, '', ext='.ts', timestamp=8)
self._create_ondisk_file(df, '', ext='.ts', timestamp=7)
self._create_ondisk_file(df, 'B', ext='.data', timestamp=6)
self._create_ondisk_file(df, 'A', ext='.data', timestamp=5)
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
try:
df.open()
except DiskFileDeleted as d:
self.assertEquals(d.timestamp, normalize_timestamp(8))
else:
self.fail("Expected DiskFileDeleted exception")
def test_ondisk_search_loop_meta_data_ts(self):
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
self._create_ondisk_file(df, '', ext='.meta', timestamp=10)
self._create_ondisk_file(df, '', ext='.meta', timestamp=9)
self._create_ondisk_file(df, 'B', ext='.data', timestamp=8)
self._create_ondisk_file(df, 'A', ext='.data', timestamp=7)
self._create_ondisk_file(df, '', ext='.ts', timestamp=6)
self._create_ondisk_file(df, '', ext='.ts', timestamp=5)
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
with df.open():
self.assertTrue('X-Timestamp' in df._metadata)
self.assertEquals(df._metadata['X-Timestamp'],
normalize_timestamp(10))
self.assertTrue('deleted' not in df._metadata)
def test_ondisk_search_loop_data_meta_ts(self):
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
self._create_ondisk_file(df, 'B', ext='.data', timestamp=10)
self._create_ondisk_file(df, 'A', ext='.data', timestamp=9)
self._create_ondisk_file(df, '', ext='.ts', timestamp=8)
self._create_ondisk_file(df, '', ext='.ts', timestamp=7)
self._create_ondisk_file(df, '', ext='.meta', timestamp=6)
self._create_ondisk_file(df, '', ext='.meta', timestamp=5)
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
with df.open():
self.assertTrue('X-Timestamp' in df._metadata)
self.assertEquals(df._metadata['X-Timestamp'],
normalize_timestamp(10))
self.assertTrue('deleted' not in df._metadata)
def test_ondisk_search_loop_wayward_files_ignored(self):
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
self._create_ondisk_file(df, 'X', ext='.bar', timestamp=11)
self._create_ondisk_file(df, 'B', ext='.data', timestamp=10)
self._create_ondisk_file(df, 'A', ext='.data', timestamp=9)
self._create_ondisk_file(df, '', ext='.ts', timestamp=8)
self._create_ondisk_file(df, '', ext='.ts', timestamp=7)
self._create_ondisk_file(df, '', ext='.meta', timestamp=6)
self._create_ondisk_file(df, '', ext='.meta', timestamp=5)
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
with df.open():
self.assertTrue('X-Timestamp' in df._metadata)
self.assertEquals(df._metadata['X-Timestamp'],
normalize_timestamp(10))
self.assertTrue('deleted' not in df._metadata)
def test_ondisk_search_loop_listdir_error(self):
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
def mock_listdir_exp(*args, **kwargs):
raise OSError(errno.EACCES, os.strerror(errno.EACCES))
with mock.patch("os.listdir", mock_listdir_exp):
self._create_ondisk_file(df, 'X', ext='.bar', timestamp=11)
self._create_ondisk_file(df, 'B', ext='.data', timestamp=10)
self._create_ondisk_file(df, 'A', ext='.data', timestamp=9)
self._create_ondisk_file(df, '', ext='.ts', timestamp=8)
self._create_ondisk_file(df, '', ext='.ts', timestamp=7)
self._create_ondisk_file(df, '', ext='.meta', timestamp=6)
self._create_ondisk_file(df, '', ext='.meta', timestamp=5)
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
self.assertRaises(DiskFileError, df.open)
def test_exception_in_handle_close_quarantine(self):
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
df = self._get_open_disk_file()
def blow_up():
raise Exception('a very special error')
DiskFile API, with reference implementation Refactor on-disk knowledge out of the object server by pushing the async update pickle creation to the new DiskFileManager class (name is not the best, so suggestions welcome), along with the REPLICATOR method logic. We also move the mount checking and thread pool storage to the new ondisk.Devices object, which then also becomes the new home of the audit_location_generator method. For the object server, a new setup() method is now called at the end of the controller's construction, and the _diskfile() method has been renamed to get_diskfile(), to allow implementation specific behavior. We then hide the need for the REST API layer to know how and where quarantining needs to be performed. There are now two places it is checked internally, on open() where we verify the content-length, name, and x-timestamp metadata, and in the reader on close where the etag metadata is checked if the entire file was read. We add a reader class to allow implementations to isolate the WSGI handling code for that specific environment (it is used no-where else in the REST APIs). This simplifies the caller's code to just use a "with" statement once open to avoid multiple points where close needs to be called. For a full historical comparison, including the usage patterns see: https://gist.github.com/portante/5488238 (as of master, 2b639f5, Merge "Fix 500 from account-quota This Commit middleware") --------------------------------+------------------------------------ DiskFileManager(conf) Methods: .pickle_async_update() .get_diskfile() .get_hashes() Attributes: .devices .logger .disk_chunk_size .keep_cache_size .bytes_per_sync DiskFile(a,c,o,keep_data_fp=) DiskFile(a,c,o) Methods: Methods: *.__iter__() .close(verify_file=) .is_deleted() .is_expired() .quarantine() .get_data_file_size() .open() .read_metadata() .create() .create() .write_metadata() .delete() .delete() Attributes: Attributes: .quarantined_dir .keep_cache .metadata *DiskFileReader() Methods: .__iter__() .close() Attributes: +.was_quarantined DiskWriter() DiskFileWriter() Methods: Methods: .write() .write() .put() .put() * Note that the DiskFile class * Note that the DiskReader() object implements all the methods returned by the necessary for a WSGI app DiskFileOpened.reader() method iterator implements all the methods necessary for a WSGI app iterator + Note that if the auditor is refactored to not use the DiskFile class, see https://review.openstack.org/44787 then we don't need the was_quarantined attribute A reference "in-memory" object server implementation of a backend DiskFile class in swift/obj/mem_server.py and swift/obj/mem_diskfile.py. One can also reference https://github.com/portante/gluster-swift/commits/diskfile for the proposed integration with the gluster-swift code based on these changes. Change-Id: I44e153fdb405a5743e9c05349008f94136764916 Signed-off-by: Peter Portante <peter.portante@redhat.com>
2013-09-12 19:51:18 -04:00
reader = df.reader()
reader._handle_close_quarantine = blow_up
for _ in reader:
pass
reader.close()
log_lines = df._logger.get_lines_for_level('error')
self.assert_('a very special error' in log_lines[-1])
Object replication ssync (an rsync alternative) For this commit, ssync is just a direct replacement for how we use rsync. Assuming we switch over to ssync completely someday and drop rsync, we will then be able to improve the algorithms even further (removing local objects as we successfully transfer each one rather than waiting for whole partitions, using an index.db with hash-trees, etc., etc.) For easier review, this commit can be thought of in distinct parts: 1) New global_conf_callback functionality for allowing services to perform setup code before workers, etc. are launched. (This is then used by ssync in the object server to create a cross-worker semaphore to restrict concurrent incoming replication.) 2) A bit of shifting of items up from object server and replicator to diskfile or DEFAULT conf sections for better sharing of the same settings. conn_timeout, node_timeout, client_timeout, network_chunk_size, disk_chunk_size. 3) Modifications to the object server and replicator to optionally use ssync in place of rsync. This is done in a generic enough way that switching to FutureSync should be easy someday. 4) The biggest part, and (at least for now) completely optional part, are the new ssync_sender and ssync_receiver files. Nice and isolated for easier testing and visibility into test coverage, etc. All the usual logging, statsd, recon, etc. instrumentation is still there when using ssync, just as it is when using rsync. Beyond the essential error and exceptional condition logging, I have not added any additional instrumentation at this time. Unless there is something someone finds super pressing to have added to the logging, I think such additions would be better as separate change reviews. FOR NOW, IT IS NOT RECOMMENDED TO USE SSYNC ON PRODUCTION CLUSTERS. Some of us will be in a limited fashion to look for any subtle issues, tuning, etc. but generally ssync is an experimental feature. In its current implementation it is probably going to be a bit slower than rsync, but if all goes according to plan it will end up much faster. There are no comparisions yet between ssync and rsync other than some raw virtual machine testing I've done to show it should compete well enough once we can put it in use in the real world. If you Tweet, Google+, or whatever, be sure to indicate it's experimental. It'd be best to keep it out of deployment guides, howtos, etc. until we all figure out if we like it, find it to be stable, etc. Change-Id: If003dcc6f4109e2d2a42f4873a0779110fff16d6
2013-08-28 16:10:43 +00:00
def test_get_diskfile_from_hash_dev_path_fail(self):
self.df_mgr.get_dev_path = mock.MagicMock(return_value=None)
with nested(
mock.patch('swift.obj.diskfile.DiskFile'),
mock.patch('swift.obj.diskfile.hash_cleanup_listdir'),
mock.patch('swift.obj.diskfile.read_metadata')) as \
(dfclass, hclistdir, readmeta):
hclistdir.return_value = ['1381679759.90941.data']
readmeta.return_value = {'name': '/a/c/o'}
self.assertRaises(
DiskFileDeviceUnavailable,
self.df_mgr.get_diskfile_from_hash,
'dev', '9', '9a7175077c01a23ade5956b8a2bba900')
def test_get_diskfile_from_hash_not_dir(self):
self.df_mgr.get_dev_path = mock.MagicMock(return_value='/srv/dev/')
with nested(
mock.patch('swift.obj.diskfile.DiskFile'),
mock.patch('swift.obj.diskfile.hash_cleanup_listdir'),
mock.patch('swift.obj.diskfile.read_metadata'),
mock.patch('swift.obj.diskfile.quarantine_renamer')) as \
(dfclass, hclistdir, readmeta, quarantine_renamer):
osexc = OSError()
osexc.errno = errno.ENOTDIR
hclistdir.side_effect = osexc
readmeta.return_value = {'name': '/a/c/o'}
self.assertRaises(
DiskFileNotExist,
self.df_mgr.get_diskfile_from_hash,
'dev', '9', '9a7175077c01a23ade5956b8a2bba900')
quarantine_renamer.assert_called_once_with(
'/srv/dev/',
'/srv/dev/objects/9/900/9a7175077c01a23ade5956b8a2bba900')
def test_get_diskfile_from_hash_no_dir(self):
self.df_mgr.get_dev_path = mock.MagicMock(return_value='/srv/dev/')
with nested(
mock.patch('swift.obj.diskfile.DiskFile'),
mock.patch('swift.obj.diskfile.hash_cleanup_listdir'),
mock.patch('swift.obj.diskfile.read_metadata')) as \
(dfclass, hclistdir, readmeta):
osexc = OSError()
osexc.errno = errno.ENOENT
hclistdir.side_effect = osexc
readmeta.return_value = {'name': '/a/c/o'}
self.assertRaises(
DiskFileNotExist,
self.df_mgr.get_diskfile_from_hash,
'dev', '9', '9a7175077c01a23ade5956b8a2bba900')
def test_get_diskfile_from_hash_other_oserror(self):
self.df_mgr.get_dev_path = mock.MagicMock(return_value='/srv/dev/')
with nested(
mock.patch('swift.obj.diskfile.DiskFile'),
mock.patch('swift.obj.diskfile.hash_cleanup_listdir'),
mock.patch('swift.obj.diskfile.read_metadata')) as \
(dfclass, hclistdir, readmeta):
osexc = OSError()
hclistdir.side_effect = osexc
readmeta.return_value = {'name': '/a/c/o'}
self.assertRaises(
OSError,
self.df_mgr.get_diskfile_from_hash,
'dev', '9', '9a7175077c01a23ade5956b8a2bba900')
def test_get_diskfile_from_hash_no_actual_files(self):
self.df_mgr.get_dev_path = mock.MagicMock(return_value='/srv/dev/')
with nested(
mock.patch('swift.obj.diskfile.DiskFile'),
mock.patch('swift.obj.diskfile.hash_cleanup_listdir'),
mock.patch('swift.obj.diskfile.read_metadata')) as \
(dfclass, hclistdir, readmeta):
hclistdir.return_value = []
readmeta.return_value = {'name': '/a/c/o'}
self.assertRaises(
DiskFileNotExist,
self.df_mgr.get_diskfile_from_hash,
'dev', '9', '9a7175077c01a23ade5956b8a2bba900')
def test_get_diskfile_from_hash_read_metadata_problem(self):
self.df_mgr.get_dev_path = mock.MagicMock(return_value='/srv/dev/')
with nested(
mock.patch('swift.obj.diskfile.DiskFile'),
mock.patch('swift.obj.diskfile.hash_cleanup_listdir'),
mock.patch('swift.obj.diskfile.read_metadata')) as \
(dfclass, hclistdir, readmeta):
hclistdir.return_value = ['1381679759.90941.data']
readmeta.side_effect = EOFError()
self.assertRaises(
DiskFileNotExist,
self.df_mgr.get_diskfile_from_hash,
'dev', '9', '9a7175077c01a23ade5956b8a2bba900')
def test_get_diskfile_from_hash_no_meta_name(self):
self.df_mgr.get_dev_path = mock.MagicMock(return_value='/srv/dev/')
with nested(
mock.patch('swift.obj.diskfile.DiskFile'),
mock.patch('swift.obj.diskfile.hash_cleanup_listdir'),
mock.patch('swift.obj.diskfile.read_metadata')) as \
(dfclass, hclistdir, readmeta):
hclistdir.return_value = ['1381679759.90941.data']
readmeta.return_value = {}
try:
self.df_mgr.get_diskfile_from_hash(
'dev', '9', '9a7175077c01a23ade5956b8a2bba900')
except DiskFileNotExist as err:
exc = err
self.assertEqual(str(exc), '')
def test_get_diskfile_from_hash_bad_meta_name(self):
self.df_mgr.get_dev_path = mock.MagicMock(return_value='/srv/dev/')
with nested(
mock.patch('swift.obj.diskfile.DiskFile'),
mock.patch('swift.obj.diskfile.hash_cleanup_listdir'),
mock.patch('swift.obj.diskfile.read_metadata')) as \
(dfclass, hclistdir, readmeta):
hclistdir.return_value = ['1381679759.90941.data']
readmeta.return_value = {'name': 'bad'}
try:
self.df_mgr.get_diskfile_from_hash(
'dev', '9', '9a7175077c01a23ade5956b8a2bba900')
except DiskFileNotExist as err:
exc = err
self.assertEqual(str(exc), '')
def test_get_diskfile_from_hash(self):
self.df_mgr.get_dev_path = mock.MagicMock(return_value='/srv/dev/')
with nested(
mock.patch('swift.obj.diskfile.DiskFile'),
mock.patch('swift.obj.diskfile.hash_cleanup_listdir'),
mock.patch('swift.obj.diskfile.read_metadata')) as \
(dfclass, hclistdir, readmeta):
hclistdir.return_value = ['1381679759.90941.data']
readmeta.return_value = {'name': '/a/c/o'}
self.df_mgr.get_diskfile_from_hash(
'dev', '9', '9a7175077c01a23ade5956b8a2bba900')
dfclass.assert_called_once_with(
self.df_mgr, '/srv/dev/', self.df_mgr.threadpools['dev'], '9',
'a', 'c', 'o')
hclistdir.assert_called_once_with(
'/srv/dev/objects/9/900/9a7175077c01a23ade5956b8a2bba900',
604800)
readmeta.assert_called_once_with(
'/srv/dev/objects/9/900/9a7175077c01a23ade5956b8a2bba900/'
'1381679759.90941.data')
def test_listdir_enoent(self):
oserror = OSError()
oserror.errno = errno.ENOENT
self.df_mgr.logger.error = mock.MagicMock()
with mock.patch('os.listdir', side_effect=oserror):
self.assertEqual(self.df_mgr._listdir('path'), [])
self.assertEqual(self.df_mgr.logger.error.mock_calls, [])
def test_listdir_other_oserror(self):
oserror = OSError()
self.df_mgr.logger.error = mock.MagicMock()
with mock.patch('os.listdir', side_effect=oserror):
self.assertEqual(self.df_mgr._listdir('path'), [])
self.df_mgr.logger.error.assert_called_once_with(
'ERROR: Skipping %r due to error with listdir attempt: %s',
'path', oserror)
def test_listdir(self):
self.df_mgr.logger.error = mock.MagicMock()
with mock.patch('os.listdir', return_value=['abc', 'def']):
self.assertEqual(self.df_mgr._listdir('path'), ['abc', 'def'])
self.assertEqual(self.df_mgr.logger.error.mock_calls, [])
def test_yield_suffixes_dev_path_fail(self):
self.df_mgr.get_dev_path = mock.MagicMock(return_value=None)
exc = None
try:
list(self.df_mgr.yield_suffixes('dev', '9'))
except DiskFileDeviceUnavailable as err:
exc = err
self.assertEqual(str(exc), '')
def test_yield_suffixes(self):
self.df_mgr._listdir = mock.MagicMock(return_value=[
'abc', 'def', 'ghi', 'abcd', '012'])
self.assertEqual(
list(self.df_mgr.yield_suffixes('dev', '9')),
[(self.testdir + '/dev/objects/9/abc', 'abc'),
(self.testdir + '/dev/objects/9/def', 'def'),
(self.testdir + '/dev/objects/9/012', '012')])
def test_yield_hashes_dev_path_fail(self):
self.df_mgr.get_dev_path = mock.MagicMock(return_value=None)
exc = None
try:
list(self.df_mgr.yield_hashes('dev', '9'))
except DiskFileDeviceUnavailable as err:
exc = err
self.assertEqual(str(exc), '')
def test_yield_hashes_empty(self):
def _listdir(path):
return []
with mock.patch('os.listdir', _listdir):
self.assertEqual(list(self.df_mgr.yield_hashes('dev', '9')), [])
def test_yield_hashes_empty_suffixes(self):
def _listdir(path):
return []
with mock.patch('os.listdir', _listdir):
self.assertEqual(
list(self.df_mgr.yield_hashes('dev', '9', suffixes=['456'])),
[])
def test_yield_hashes(self):
fresh_ts = normalize_timestamp(time() - 10)
fresher_ts = normalize_timestamp(time() - 1)
def _listdir(path):
if path.endswith('/dev/objects/9'):
return ['abc', '456', 'def']
elif path.endswith('/dev/objects/9/abc'):
return ['9373a92d072897b136b3fc06595b4abc']
elif path.endswith(
'/dev/objects/9/abc/9373a92d072897b136b3fc06595b4abc'):
return [fresh_ts + '.ts']
elif path.endswith('/dev/objects/9/456'):
return ['9373a92d072897b136b3fc06595b0456',
'9373a92d072897b136b3fc06595b7456']
elif path.endswith(
'/dev/objects/9/456/9373a92d072897b136b3fc06595b0456'):
return ['1383180000.12345.data']
elif path.endswith(
'/dev/objects/9/456/9373a92d072897b136b3fc06595b7456'):
return [fresh_ts + '.ts',
fresher_ts + '.data']
elif path.endswith('/dev/objects/9/def'):
return []
else:
raise Exception('Unexpected listdir of %r' % path)
with nested(
mock.patch('os.listdir', _listdir),
mock.patch('os.unlink')):
self.assertEqual(
list(self.df_mgr.yield_hashes('dev', '9')),
[(self.testdir +
'/dev/objects/9/abc/9373a92d072897b136b3fc06595b4abc',
'9373a92d072897b136b3fc06595b4abc', fresh_ts),
(self.testdir +
'/dev/objects/9/456/9373a92d072897b136b3fc06595b0456',
'9373a92d072897b136b3fc06595b0456', '1383180000.12345'),
(self.testdir +
'/dev/objects/9/456/9373a92d072897b136b3fc06595b7456',
'9373a92d072897b136b3fc06595b7456', fresher_ts)])
def test_yield_hashes_suffixes(self):
fresh_ts = normalize_timestamp(time() - 10)
fresher_ts = normalize_timestamp(time() - 1)
def _listdir(path):
if path.endswith('/dev/objects/9'):
return ['abc', '456', 'def']
elif path.endswith('/dev/objects/9/abc'):
return ['9373a92d072897b136b3fc06595b4abc']
elif path.endswith(
'/dev/objects/9/abc/9373a92d072897b136b3fc06595b4abc'):
return [fresh_ts + '.ts']
elif path.endswith('/dev/objects/9/456'):
return ['9373a92d072897b136b3fc06595b0456',
'9373a92d072897b136b3fc06595b7456']
elif path.endswith(
'/dev/objects/9/456/9373a92d072897b136b3fc06595b0456'):
return ['1383180000.12345.data']
elif path.endswith(
'/dev/objects/9/456/9373a92d072897b136b3fc06595b7456'):
return [fresh_ts + '.ts',
fresher_ts + '.data']
elif path.endswith('/dev/objects/9/def'):
return []
else:
raise Exception('Unexpected listdir of %r' % path)
with nested(
mock.patch('os.listdir', _listdir),
mock.patch('os.unlink')):
self.assertEqual(
list(self.df_mgr.yield_hashes(
'dev', '9', suffixes=['456'])),
[(self.testdir +
'/dev/objects/9/456/9373a92d072897b136b3fc06595b0456',
'9373a92d072897b136b3fc06595b0456', '1383180000.12345'),
(self.testdir +
'/dev/objects/9/456/9373a92d072897b136b3fc06595b7456',
'9373a92d072897b136b3fc06595b7456', fresher_ts)])
def test_diskfile_names(self):
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
self.assertEqual(df.account, 'a')
self.assertEqual(df.container, 'c')
self.assertEqual(df.obj, 'o')
def test_diskfile_content_length_not_open(self):
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
exc = None
try:
df.content_length
except DiskFileNotOpen as err:
exc = err
self.assertEqual(str(exc), '')
def test_diskfile_content_length_deleted(self):
df = self._get_open_disk_file()
ts = time()
df.delete(ts)
exp_name = '%s.ts' % str(normalize_timestamp(ts))
dl = os.listdir(df._datadir)
self.assertEquals(len(dl), 1)
self.assertTrue(exp_name in set(dl))
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
exc = None
try:
with df.open():
df.content_length
except DiskFileDeleted as err:
exc = err
self.assertEqual(str(exc), '')
def test_diskfile_content_length(self):
self._get_open_disk_file()
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
with df.open():
self.assertEqual(df.content_length, 1024)
def test_diskfile_timestamp_not_open(self):
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
exc = None
try:
df.timestamp
except DiskFileNotOpen as err:
exc = err
self.assertEqual(str(exc), '')
def test_diskfile_timestamp_deleted(self):
df = self._get_open_disk_file()
ts = time()
df.delete(ts)
exp_name = '%s.ts' % str(normalize_timestamp(ts))
dl = os.listdir(df._datadir)
self.assertEquals(len(dl), 1)
self.assertTrue(exp_name in set(dl))
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
exc = None
try:
with df.open():
df.timestamp
except DiskFileDeleted as err:
exc = err
self.assertEqual(str(exc), '')
def test_diskfile_timestamp(self):
self._get_open_disk_file(ts='1383181759.12345')
df = self.df_mgr.get_diskfile('sda1', '0', 'a', 'c', 'o')
with df.open():
self.assertEqual(df.timestamp, '1383181759.12345')
def test_error_in_hash_cleanup_listdir(self):
def mock_hcl(*args, **kwargs):
raise OSError()
df = self._get_open_disk_file()
ts = time()
with mock.patch("swift.obj.diskfile.hash_cleanup_listdir",
mock_hcl):
try:
df.delete(ts)
except OSError:
self.fail("OSError raised when it should have been swallowed")
exp_name = '%s.ts' % str(normalize_timestamp(ts))
dl = os.listdir(df._datadir)
self.assertEquals(len(dl), 2)
self.assertTrue(exp_name in set(dl))
Object replication ssync (an rsync alternative) For this commit, ssync is just a direct replacement for how we use rsync. Assuming we switch over to ssync completely someday and drop rsync, we will then be able to improve the algorithms even further (removing local objects as we successfully transfer each one rather than waiting for whole partitions, using an index.db with hash-trees, etc., etc.) For easier review, this commit can be thought of in distinct parts: 1) New global_conf_callback functionality for allowing services to perform setup code before workers, etc. are launched. (This is then used by ssync in the object server to create a cross-worker semaphore to restrict concurrent incoming replication.) 2) A bit of shifting of items up from object server and replicator to diskfile or DEFAULT conf sections for better sharing of the same settings. conn_timeout, node_timeout, client_timeout, network_chunk_size, disk_chunk_size. 3) Modifications to the object server and replicator to optionally use ssync in place of rsync. This is done in a generic enough way that switching to FutureSync should be easy someday. 4) The biggest part, and (at least for now) completely optional part, are the new ssync_sender and ssync_receiver files. Nice and isolated for easier testing and visibility into test coverage, etc. All the usual logging, statsd, recon, etc. instrumentation is still there when using ssync, just as it is when using rsync. Beyond the essential error and exceptional condition logging, I have not added any additional instrumentation at this time. Unless there is something someone finds super pressing to have added to the logging, I think such additions would be better as separate change reviews. FOR NOW, IT IS NOT RECOMMENDED TO USE SSYNC ON PRODUCTION CLUSTERS. Some of us will be in a limited fashion to look for any subtle issues, tuning, etc. but generally ssync is an experimental feature. In its current implementation it is probably going to be a bit slower than rsync, but if all goes according to plan it will end up much faster. There are no comparisions yet between ssync and rsync other than some raw virtual machine testing I've done to show it should compete well enough once we can put it in use in the real world. If you Tweet, Google+, or whatever, be sure to indicate it's experimental. It'd be best to keep it out of deployment guides, howtos, etc. until we all figure out if we like it, find it to be stable, etc. Change-Id: If003dcc6f4109e2d2a42f4873a0779110fff16d6
2013-08-28 16:10:43 +00:00
if __name__ == '__main__':
unittest.main()