swift/test/unit/obj/test_replicator.py

812 lines
35 KiB
Python
Raw Normal View History

# Copyright (c) 2010-2012 OpenStack Foundation
2010-07-12 17:03:45 -05:00
#
# 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.
import unittest
import os
import mock
2010-07-12 17:03:45 -05:00
from gzip import GzipFile
from shutil import rmtree
import cPickle as pickle
2010-10-29 15:26:35 -07:00
import time
2010-11-16 11:06:39 -08:00
import tempfile
from contextlib import contextmanager, closing
2010-07-12 17:03:45 -05:00
from eventlet.green import subprocess
from eventlet import Timeout, tpool
from test.unit import FakeLogger, patch_policies
from swift.common import utils
from swift.common.utils import hash_path, mkdirs, normalize_timestamp
2010-07-12 17:03:45 -05:00
from swift.common import ring
from swift.obj import diskfile, replicator as object_replicator
from swift.common.storage_policy import StoragePolicy, POLICIES
2010-07-12 17:03:45 -05:00
2010-07-12 17:03:45 -05:00
def _ips():
return ['127.0.0.0']
2010-07-12 17:03:45 -05:00
object_replicator.whataremyips = _ips
2010-10-29 15:26:35 -07:00
def mock_http_connect(status):
class FakeConn(object):
def __init__(self, status, *args, **kwargs):
self.status = status
self.reason = 'Fake'
self.host = args[0]
self.port = args[1]
self.method = args[4]
self.path = args[5]
self.with_exc = False
self.headers = kwargs.get('headers', {})
2010-10-29 15:26:35 -07:00
def getresponse(self):
if self.with_exc:
raise Exception('test')
return self
def getheader(self, header):
return self.headers[header]
def read(self, amt=None):
return pickle.dumps({})
def close(self):
return
return lambda *args, **kwargs: FakeConn(status, *args, **kwargs)
process_errors = []
2010-07-12 17:03:45 -05:00
class MockProcess(object):
ret_code = None
ret_log = None
2010-10-29 15:26:35 -07:00
check_args = None
2010-07-12 17:03:45 -05:00
class Stream(object):
2010-07-12 17:03:45 -05:00
def read(self):
return MockProcess.ret_log.next()
def __init__(self, *args, **kwargs):
2010-10-29 15:26:35 -07:00
targs = MockProcess.check_args.next()
for targ in targs:
if targ not in args[0]:
process_errors.append("Invalid: %s not in %s" % (targ,
args))
2010-07-12 17:03:45 -05:00
self.stdout = self.Stream()
def wait(self):
return self.ret_code.next()
2010-07-12 17:03:45 -05:00
@contextmanager
def _mock_process(ret):
orig_process = subprocess.Popen
MockProcess.ret_code = (i[0] for i in ret)
MockProcess.ret_log = (i[1] for i in ret)
2010-10-29 15:26:35 -07:00
MockProcess.check_args = (i[2] for i in ret)
2010-07-12 17:03:45 -05:00
object_replicator.subprocess.Popen = MockProcess
yield
object_replicator.subprocess.Popen = orig_process
def _create_test_rings(path):
2010-07-12 17:03:45 -05:00
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],
]
2010-07-12 17:03:45 -05:00
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},
]
2010-07-12 17:03:45 -05:00
intended_part_shift = 30
with closing(GzipFile(testgz, 'wb')) as f:
pickle.dump(
ring.RingData(intended_replica2part2dev_id,
intended_devs, intended_part_shift),
f)
testgz = os.path.join(path, 'object-1.ring.gz')
with closing(GzipFile(testgz, 'wb')) as f:
pickle.dump(
ring.RingData(intended_replica2part2dev_id,
intended_devs, intended_part_shift),
f)
return
2010-07-12 17:03:45 -05:00
@patch_policies([StoragePolicy(0, 'zero', False),
StoragePolicy(1, 'one', True)])
2010-07-12 17:03:45 -05:00
class TestObjectReplicator(unittest.TestCase):
def setUp(self):
utils.HASH_PATH_SUFFIX = 'endcap'
utils.HASH_PATH_PREFIX = ''
2010-07-12 17:03:45 -05:00
# Setup a test ring (stolen from common/test_ring.py)
2010-11-16 11:06:39 -08:00
self.testdir = tempfile.mkdtemp()
2010-07-12 17:03:45 -05:00
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',
diskfile.get_data_dir(0))
self.objects_1 = os.path.join(self.devices, 'sda',
diskfile.get_data_dir(1))
2010-07-12 17:03:45 -05:00
os.mkdir(self.objects)
os.mkdir(self.objects_1)
2010-10-29 15:26:35 -07:00
self.parts = {}
self.parts_1 = {}
for part in ['0', '1', '2', '3']:
2010-10-29 15:26:35 -07:00
self.parts[part] = os.path.join(self.objects, part)
2010-07-12 17:03:45 -05:00
os.mkdir(os.path.join(self.objects, part))
self.parts_1[part] = os.path.join(self.objects_1, part)
os.mkdir(os.path.join(self.objects_1, part))
_create_test_rings(self.testdir)
2010-07-12 17:03:45 -05:00
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.replicator = object_replicator.ObjectReplicator(self.conf)
self.replicator.logger = FakeLogger()
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,
self.replicator.logger)
2010-07-12 17:03:45 -05:00
2010-10-29 15:26:35 -07:00
def tearDown(self):
rmtree(self.testdir, ignore_errors=1)
def test_run_once(self):
conf = dict(swift_dir=self.testdir, devices=self.devices,
mount_check='false', timeout='300', stats_interval='1')
replicator = object_replicator.ObjectReplicator(conf)
was_connector = object_replicator.http_connect
2010-10-29 15:26:35 -07:00
object_replicator.http_connect = mock_http_connect(200)
cur_part = '0'
df = self.df_mgr.get_diskfile('sda', cur_part, 'a', 'c', 'o',
policy_idx=0)
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
mkdirs(df._datadir)
f = open(os.path.join(df._datadir,
2010-10-29 15:26:35 -07:00
normalize_timestamp(time.time()) + '.data'),
'wb')
f.write('1234567890')
f.close()
ohash = hash_path('a', 'c', 'o')
data_dir = ohash[-3:]
whole_path_from = os.path.join(self.objects, cur_part, data_dir)
process_arg_checker = []
ring = replicator.get_object_ring(0)
2010-10-29 15:26:35 -07:00
nodes = [node for node in
ring.get_part_nodes(int(cur_part))
if node['ip'] not in _ips()]
2010-10-29 15:26:35 -07:00
for node in nodes:
rsync_mod = '%s::object/sda/objects/%s' % (node['ip'], cur_part)
process_arg_checker.append(
(0, '', ['rsync', whole_path_from, rsync_mod]))
2010-10-29 15:26:35 -07:00
with _mock_process(process_arg_checker):
replicator.run_once()
self.assertFalse(process_errors)
object_replicator.http_connect = was_connector
2010-10-29 15:26:35 -07:00
# policy 1
def test_run_once_1(self):
conf = dict(swift_dir=self.testdir, devices=self.devices,
mount_check='false', timeout='300', stats_interval='1')
replicator = object_replicator.ObjectReplicator(conf)
was_connector = object_replicator.http_connect
object_replicator.http_connect = mock_http_connect(200)
cur_part = '0'
df = self.df_mgr.get_diskfile('sda', cur_part, 'a', 'c', 'o',
policy_idx=1)
mkdirs(df._datadir)
f = open(os.path.join(df._datadir,
normalize_timestamp(time.time()) + '.data'),
'wb')
f.write('1234567890')
f.close()
ohash = hash_path('a', 'c', 'o')
data_dir = ohash[-3:]
whole_path_from = os.path.join(self.objects_1, cur_part, data_dir)
process_arg_checker = []
ring = replicator.get_object_ring(1)
nodes = [node for node in
ring.get_part_nodes(int(cur_part))
if node['ip'] not in _ips()]
for node in nodes:
rsync_mod = '%s::object/sda/objects-1/%s' % (node['ip'], cur_part)
process_arg_checker.append(
(0, '', ['rsync', whole_path_from, rsync_mod]))
with _mock_process(process_arg_checker):
replicator.run_once()
self.assertFalse(process_errors)
object_replicator.http_connect = was_connector
def test_check_ring(self):
for pol in POLICIES:
obj_ring = self.replicator.get_object_ring(pol.idx)
self.assertTrue(self.replicator.check_ring(obj_ring))
orig_check = self.replicator.next_check
self.replicator.next_check = orig_check - 30
self.assertTrue(self.replicator.check_ring(obj_ring))
self.replicator.next_check = orig_check
orig_ring_time = obj_ring._mtime
obj_ring._mtime = orig_ring_time - 30
self.assertTrue(self.replicator.check_ring(obj_ring))
self.replicator.next_check = orig_check - 30
self.assertFalse(self.replicator.check_ring(obj_ring))
def test_collect_jobs_mkdirs_error(self):
def blowup_mkdirs(path):
raise OSError('Ow!')
mkdirs_orig = object_replicator.mkdirs
try:
rmtree(self.objects, ignore_errors=1)
object_replicator.mkdirs = blowup_mkdirs
self.replicator.collect_jobs()
self.assertTrue('exception' in self.replicator.logger.log_dict)
self.assertEquals(
len(self.replicator.logger.log_dict['exception']), 1)
exc_args, exc_kwargs, exc_str = \
self.replicator.logger.log_dict['exception'][0]
self.assertEquals(len(exc_args), 1)
self.assertTrue(exc_args[0].startswith('ERROR creating '))
self.assertEquals(exc_kwargs, {})
self.assertEquals(exc_str, 'Ow!')
finally:
object_replicator.mkdirs = mkdirs_orig
def test_collect_jobs(self):
jobs = self.replicator.collect_jobs()
jobs_to_delete = [j for j in jobs if j['delete']]
jobs_by_pol_part = {}
for job in jobs:
jobs_by_pol_part[str(job['policy_idx']) + job['partition']] = job
self.assertEquals(len(jobs_to_delete), 2)
self.assertTrue('1', jobs_to_delete[0]['partition'])
self.assertEquals(
[node['id'] for node in jobs_by_pol_part['00']['nodes']], [1, 2])
self.assertEquals(
[node['id'] for node in jobs_by_pol_part['01']['nodes']],
[1, 2, 3])
self.assertEquals(
[node['id'] for node in jobs_by_pol_part['02']['nodes']], [2, 3])
self.assertEquals(
[node['id'] for node in jobs_by_pol_part['03']['nodes']], [3, 1])
self.assertEquals(
[node['id'] for node in jobs_by_pol_part['10']['nodes']], [1, 2])
self.assertEquals(
[node['id'] for node in jobs_by_pol_part['11']['nodes']],
[1, 2, 3])
self.assertEquals(
[node['id'] for node in jobs_by_pol_part['12']['nodes']], [2, 3])
self.assertEquals(
[node['id'] for node in jobs_by_pol_part['13']['nodes']], [3, 1])
for part in ['00', '01', '02', '03', ]:
for node in jobs_by_pol_part[part]['nodes']:
self.assertEquals(node['device'], 'sda')
self.assertEquals(jobs_by_pol_part[part]['path'],
os.path.join(self.objects, part[1:]))
for part in ['10', '11', '12', '13', ]:
for node in jobs_by_pol_part[part]['nodes']:
self.assertEquals(node['device'], 'sda')
self.assertEquals(jobs_by_pol_part[part]['path'],
os.path.join(self.objects_1, part[1:]))
def test_collect_jobs_handoffs_first(self):
self.replicator.handoffs_first = True
jobs = self.replicator.collect_jobs()
self.assertTrue(jobs[0]['delete'])
self.assertEquals('1', jobs[0]['partition'])
def test_collect_jobs_removes_zbf(self):
"""
After running xfs_repair, a partition directory could become a
zero-byte file. If this happens, collect_jobs() should clean it up and
*not* create a job which will hit an exception as it tries to listdir()
a file.
"""
# Surprise! Partition dir 1 is actually a zero-byte-file
part_1_path = os.path.join(self.objects, '1')
rmtree(part_1_path)
with open(part_1_path, 'w'):
pass
self.assertTrue(os.path.isfile(part_1_path)) # sanity check
part_1_path_1 = os.path.join(self.objects_1, '1')
rmtree(part_1_path_1)
with open(part_1_path_1, 'w'):
pass
self.assertTrue(os.path.isfile(part_1_path_1)) # sanity check
jobs = self.replicator.collect_jobs()
jobs_to_delete = [j for j in jobs if j['delete']]
jobs_by_pol_part = {}
for job in jobs:
jobs_by_pol_part[str(job['policy_idx']) + job['partition']] = job
self.assertEquals(len(jobs_to_delete), 0)
self.assertEquals(
[node['id'] for node in jobs_by_pol_part['00']['nodes']], [1, 2])
self.assertFalse('1' in jobs_by_pol_part)
self.assertEquals(
[node['id'] for node in jobs_by_pol_part['02']['nodes']], [2, 3])
self.assertEquals(
[node['id'] for node in jobs_by_pol_part['03']['nodes']], [3, 1])
self.assertEquals(
[node['id'] for node in jobs_by_pol_part['10']['nodes']], [1, 2])
self.assertFalse('1' in jobs_by_pol_part)
self.assertEquals(
[node['id'] for node in jobs_by_pol_part['12']['nodes']], [2, 3])
self.assertEquals(
[node['id'] for node in jobs_by_pol_part['13']['nodes']], [3, 1])
for part in ['00', '02', '03']:
for node in jobs_by_pol_part[part]['nodes']:
self.assertEquals(node['device'], 'sda')
self.assertEquals(jobs_by_pol_part[part]['path'],
os.path.join(self.objects, part[1:]))
self.assertFalse(os.path.exists(part_1_path))
expected = sorted(self.replicator.logger.log_dict['warning'])
self.assertEquals(
(('Removing partition directory which was a file: %s',
part_1_path), {}), expected[1])
# policy 1
for part in ['10', '12', '13']:
for node in jobs_by_pol_part[part]['nodes']:
self.assertEquals(node['device'], 'sda')
self.assertEquals(jobs_by_pol_part[part]['path'],
os.path.join(self.objects_1, part[1:]))
self.assertFalse(os.path.exists(part_1_path_1))
self.assertEquals(
(('Removing partition directory which was a file: %s',
part_1_path_1), {}), expected[0])
def test_delete_partition(self):
with mock.patch('swift.obj.replicator.http_connect',
mock_http_connect(200)):
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', '1', 'a', 'c', 'o')
mkdirs(df._datadir)
f = open(os.path.join(df._datadir,
normalize_timestamp(time.time()) + '.data'),
'wb')
f.write('1234567890')
f.close()
ohash = hash_path('a', 'c', 'o')
data_dir = ohash[-3:]
whole_path_from = os.path.join(self.objects, '1', data_dir)
part_path = os.path.join(self.objects, '1')
self.assertTrue(os.access(part_path, os.F_OK))
ring = self.replicator.get_object_ring(0)
nodes = [node for node in
ring.get_part_nodes(1)
if node['ip'] not in _ips()]
process_arg_checker = []
for node in nodes:
rsync_mod = '%s::object/sda/objects/%s' % (node['ip'], 1)
process_arg_checker.append(
(0, '', ['rsync', whole_path_from, rsync_mod]))
with _mock_process(process_arg_checker):
self.replicator.replicate()
self.assertFalse(os.access(part_path, os.F_OK))
def test_delete_partition_1(self):
with mock.patch('swift.obj.replicator.http_connect',
mock_http_connect(200)):
df = self.df_mgr.get_diskfile('sda', '1', 'a', 'c', 'o',
policy_idx=1)
mkdirs(df._datadir)
f = open(os.path.join(df._datadir,
normalize_timestamp(time.time()) + '.data'),
'wb')
f.write('1234567890')
f.close()
ohash = hash_path('a', 'c', 'o')
data_dir = ohash[-3:]
whole_path_from = os.path.join(self.objects_1, '1', data_dir)
part_path = os.path.join(self.objects_1, '1')
self.assertTrue(os.access(part_path, os.F_OK))
ring = self.replicator.get_object_ring(1)
nodes = [node for node in
ring.get_part_nodes(1)
if node['ip'] not in _ips()]
process_arg_checker = []
for node in nodes:
rsync_mod = '%s::object/sda/objects-1/%s' % (node['ip'], 1)
process_arg_checker.append(
(0, '', ['rsync', whole_path_from, rsync_mod]))
with _mock_process(process_arg_checker):
self.replicator.replicate()
self.assertFalse(os.access(part_path, os.F_OK))
def test_delete_partition_with_failures(self):
with mock.patch('swift.obj.replicator.http_connect',
mock_http_connect(200)):
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', '1', 'a', 'c', 'o')
mkdirs(df._datadir)
f = open(os.path.join(df._datadir,
normalize_timestamp(time.time()) + '.data'),
'wb')
f.write('1234567890')
f.close()
ohash = hash_path('a', 'c', 'o')
data_dir = ohash[-3:]
whole_path_from = os.path.join(self.objects, '1', data_dir)
part_path = os.path.join(self.objects, '1')
self.assertTrue(os.access(part_path, os.F_OK))
ring = self.replicator.get_object_ring(0)
nodes = [node for node in
ring.get_part_nodes(1)
if node['ip'] not in _ips()]
process_arg_checker = []
for i, node in enumerate(nodes):
rsync_mod = '%s::object/sda/objects/%s' % (node['ip'], 1)
if i == 0:
# force one of the rsync calls to fail
ret_code = 1
else:
ret_code = 0
process_arg_checker.append(
(ret_code, '', ['rsync', whole_path_from, rsync_mod]))
with _mock_process(process_arg_checker):
self.replicator.replicate()
# The path should still exist
self.assertTrue(os.access(part_path, os.F_OK))
def test_delete_partition_with_handoff_delete(self):
with mock.patch('swift.obj.replicator.http_connect',
mock_http_connect(200)):
self.replicator.handoff_delete = 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
df = self.df_mgr.get_diskfile('sda', '1', 'a', 'c', 'o')
mkdirs(df._datadir)
f = open(os.path.join(df._datadir,
normalize_timestamp(time.time()) + '.data'),
'wb')
f.write('1234567890')
f.close()
ohash = hash_path('a', 'c', 'o')
data_dir = ohash[-3:]
whole_path_from = os.path.join(self.objects, '1', data_dir)
part_path = os.path.join(self.objects, '1')
self.assertTrue(os.access(part_path, os.F_OK))
ring = self.replicator.get_object_ring(0)
nodes = [node for node in
ring.get_part_nodes(1)
if node['ip'] not in _ips()]
process_arg_checker = []
for i, node in enumerate(nodes):
rsync_mod = '%s::object/sda/objects/%s' % (node['ip'], 1)
if i == 0:
# force one of the rsync calls to fail
ret_code = 1
else:
ret_code = 0
process_arg_checker.append(
(ret_code, '', ['rsync', whole_path_from, rsync_mod]))
with _mock_process(process_arg_checker):
self.replicator.replicate()
self.assertFalse(os.access(part_path, os.F_OK))
def test_delete_partition_with_handoff_delete_failures(self):
with mock.patch('swift.obj.replicator.http_connect',
mock_http_connect(200)):
self.replicator.handoff_delete = 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
df = self.df_mgr.get_diskfile('sda', '1', 'a', 'c', 'o')
mkdirs(df._datadir)
f = open(os.path.join(df._datadir,
normalize_timestamp(time.time()) + '.data'),
'wb')
f.write('1234567890')
f.close()
ohash = hash_path('a', 'c', 'o')
data_dir = ohash[-3:]
whole_path_from = os.path.join(self.objects, '1', data_dir)
part_path = os.path.join(self.objects, '1')
self.assertTrue(os.access(part_path, os.F_OK))
ring = self.replicator.get_object_ring(0)
nodes = [node for node in
ring.get_part_nodes(1)
if node['ip'] not in _ips()]
process_arg_checker = []
for i, node in enumerate(nodes):
rsync_mod = '%s::object/sda/objects/%s' % (node['ip'], 1)
if i in (0, 1):
# force two of the rsync calls to fail
ret_code = 1
else:
ret_code = 0
process_arg_checker.append(
(ret_code, '', ['rsync', whole_path_from, rsync_mod]))
with _mock_process(process_arg_checker):
self.replicator.replicate()
# The file should still exist
self.assertTrue(os.access(part_path, os.F_OK))
def test_delete_partition_override_params(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_path = os.path.join(self.objects, '1')
self.assertTrue(os.access(part_path, os.F_OK))
self.replicator.replicate(override_devices=['sdb'])
self.assertTrue(os.access(part_path, os.F_OK))
self.replicator.replicate(override_partitions=['9'])
self.assertTrue(os.access(part_path, os.F_OK))
self.replicator.replicate(override_devices=['sda'],
override_partitions=['1'])
self.assertFalse(os.access(part_path, os.F_OK))
def test_run_once_recover_from_failure(self):
conf = dict(swift_dir=self.testdir, devices=self.devices,
mount_check='false', timeout='300', stats_interval='1')
replicator = object_replicator.ObjectReplicator(conf)
was_connector = object_replicator.http_connect
try:
object_replicator.http_connect = mock_http_connect(200)
# Write some files into '1' and run replicate- they should be moved
# to the other partitoins and then node should get deleted.
cur_part = '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
df = self.df_mgr.get_diskfile('sda', cur_part, 'a', 'c', 'o')
mkdirs(df._datadir)
f = open(os.path.join(df._datadir,
normalize_timestamp(time.time()) + '.data'),
'wb')
f.write('1234567890')
f.close()
ohash = hash_path('a', 'c', 'o')
data_dir = ohash[-3:]
whole_path_from = os.path.join(self.objects, cur_part, data_dir)
ring = replicator.get_object_ring(0)
process_arg_checker = []
nodes = [node for node in
ring.get_part_nodes(int(cur_part))
if node['ip'] not in _ips()]
for node in nodes:
rsync_mod = '%s::object/sda/objects/%s' % (node['ip'],
cur_part)
process_arg_checker.append(
(0, '', ['rsync', whole_path_from, rsync_mod]))
self.assertTrue(os.access(os.path.join(self.objects,
'1', data_dir, ohash),
os.F_OK))
with _mock_process(process_arg_checker):
replicator.run_once()
self.assertFalse(process_errors)
for i, result in [('0', True), ('1', False),
('2', True), ('3', True)]:
self.assertEquals(os.access(
os.path.join(self.objects,
i, diskfile.HASH_FILE),
os.F_OK), result)
finally:
object_replicator.http_connect = was_connector
def test_run_once_recover_from_timeout(self):
conf = dict(swift_dir=self.testdir, devices=self.devices,
mount_check='false', timeout='300', stats_interval='1')
replicator = object_replicator.ObjectReplicator(conf)
was_connector = object_replicator.http_connect
was_get_hashes = object_replicator.get_hashes
was_execute = tpool.execute
self.get_hash_count = 0
try:
def fake_get_hashes(*args, **kwargs):
self.get_hash_count += 1
if self.get_hash_count == 3:
# raise timeout on last call to get hashes
raise Timeout()
return 2, {'abc': 'def'}
def fake_exc(tester, *args, **kwargs):
if 'Error syncing partition' in args[0]:
tester.i_failed = True
self.i_failed = False
object_replicator.http_connect = mock_http_connect(200)
object_replicator.get_hashes = fake_get_hashes
replicator.logger.exception = \
lambda *args, **kwargs: fake_exc(self, *args, **kwargs)
# Write some files into '1' and run replicate- they should be moved
As-unique-as-possible partition replica placement. This commit introduces a new algorithm for assigning partition replicas to devices. Basically, the ring builder organizes the devices into tiers (first zone, then IP/port, then device ID). When placing a replica, the ring builder looks for the emptiest device (biggest parts_wanted) in the furthest-away tier. In the case where zone-count >= replica-count, the new algorithm will give the same results as the one it replaces. Thus, no migration is needed. In the case where zone-count < replica-count, the new algorithm behaves differently from the old algorithm. The new algorithm will distribute things evenly at each tier so that the replication is as high-quality as possible, given the circumstances. The old algorithm would just crash, so again, no migration is needed. Handoffs have also been updated to use the new algorithm. When generating handoff nodes, first the ring looks for nodes in other zones, then other ips/ports, then any other drive. The first handoff nodes (the ones in other zones) will be the same as before; this commit just extends the list of handoff nodes. The proxy server and replicators have been altered to avoid looking at the ring's replica count directly. Previously, with a replica count of C, RingData.get_nodes() and RingData.get_part_nodes() would return lists of length C, so some other code used the replica count when it needed the number of nodes. If two of a partition's replicas are on the same device (e.g. with 3 replicas, 2 devices), then that assumption is no longer true. Fortunately, all the proxy server and replicators really needed was the number of nodes returned, which they already had. (Bonus: now the only code that mentions replica_count directly is in the ring and the ring builder.) Change-Id: Iba2929edfc6ece89791890d0635d4763d821a3aa
2012-04-23 10:41:44 -07:00
# to the other partitions and then node should get deleted.
cur_part = '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
df = self.df_mgr.get_diskfile('sda', cur_part, 'a', 'c', 'o')
mkdirs(df._datadir)
f = open(os.path.join(df._datadir,
normalize_timestamp(time.time()) + '.data'),
'wb')
f.write('1234567890')
f.close()
ohash = hash_path('a', 'c', 'o')
data_dir = ohash[-3:]
whole_path_from = os.path.join(self.objects, cur_part, data_dir)
process_arg_checker = []
ring = replicator.get_object_ring(0)
nodes = [node for node in
ring.get_part_nodes(int(cur_part))
if node['ip'] not in _ips()]
for node in nodes:
rsync_mod = '%s::object/sda/objects/%s' % (node['ip'],
cur_part)
process_arg_checker.append(
(0, '', ['rsync', whole_path_from, rsync_mod]))
self.assertTrue(os.access(os.path.join(self.objects,
'1', data_dir, ohash),
os.F_OK))
with _mock_process(process_arg_checker):
replicator.run_once()
self.assertFalse(process_errors)
self.assertFalse(self.i_failed)
finally:
object_replicator.http_connect = was_connector
object_replicator.get_hashes = was_get_hashes
tpool.execute = was_execute
2010-07-12 17:03:45 -05:00
def test_run(self):
2010-11-16 08:32:03 -08:00
with _mock_process([(0, '')] * 100):
with mock.patch('swift.obj.replicator.http_connect',
mock_http_connect(200)):
self.replicator.replicate()
2010-07-12 17:03:45 -05:00
def test_run_withlog(self):
2010-11-16 08:32:03 -08:00
with _mock_process([(0, "stuff in log")] * 100):
with mock.patch('swift.obj.replicator.http_connect',
mock_http_connect(200)):
self.replicator.replicate()
2010-07-12 17:03:45 -05:00
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_sync_just_calls_sync_method(self):
self.replicator.sync_method = mock.MagicMock()
self.replicator.sync('node', 'job', 'suffixes')
self.replicator.sync_method.assert_called_once_with(
'node', 'job', 'suffixes')
@mock.patch('swift.obj.replicator.tpool_reraise', autospec=True)
@mock.patch('swift.obj.replicator.http_connect', autospec=True)
def test_update(self, mock_http, mock_tpool_reraise):
def set_default(self):
self.replicator.suffix_count = 0
self.replicator.suffix_sync = 0
self.replicator.suffix_hash = 0
self.replicator.replication_count = 0
self.replicator.partition_times = []
self.headers = {'Content-Length': '0',
'user-agent': 'obj-replicator %s' % os.getpid()}
self.replicator.logger = mock_logger = mock.MagicMock()
mock_tpool_reraise.return_value = (0, {})
all_jobs = self.replicator.collect_jobs()
jobs = [job for job in all_jobs if not job['delete']]
mock_http.return_value = answer = mock.MagicMock()
answer.getresponse.return_value = resp = mock.MagicMock()
# Check uncorrect http_connect with status 507 and
# count of attempts and call args
resp.status = 507
error = '%(ip)s/%(device)s responded as unmounted'
expect = 'Error syncing partition'
for job in jobs:
set_default(self)
ring = self.replicator.get_object_ring(job['policy_idx'])
self.headers['X-Backend-Storage-Policy-Index'] = job['policy_idx']
self.replicator.update(job)
self.assertTrue(error in mock_logger.error.call_args[0][0])
self.assertTrue(expect in mock_logger.exception.call_args[0][0])
self.assertEquals(len(self.replicator.partition_times), 1)
self.assertEquals(mock_http.call_count, len(ring._devs) - 1)
reqs = []
for node in job['nodes']:
reqs.append(mock.call(node['ip'], node['port'], node['device'],
job['partition'], 'REPLICATE', '',
headers=self.headers))
if job['partition'] == '0':
self.assertEquals(self.replicator.suffix_hash, 0)
mock_http.assert_has_calls(reqs, any_order=True)
mock_http.reset_mock()
mock_logger.reset_mock()
# Check uncorrect http_connect with status 400 != HTTP_OK
resp.status = 400
error = 'Invalid response %(resp)s from %(ip)s'
for job in jobs:
set_default(self)
self.replicator.update(job)
self.assertTrue(error in mock_logger.error.call_args[0][0])
self.assertEquals(len(self.replicator.partition_times), 1)
mock_logger.reset_mock()
# Check successful http_connection and exception with
# uncorrect pickle.loads(resp.read())
resp.status = 200
expect = 'Error syncing with node:'
for job in jobs:
set_default(self)
self.replicator.update(job)
self.assertTrue(expect in mock_logger.exception.call_args[0][0])
self.assertEquals(len(self.replicator.partition_times), 1)
mock_logger.reset_mock()
# Check successful http_connection and correct
# pickle.loads(resp.read()) for non local node
resp.status = 200
local_job = None
resp.read.return_value = pickle.dumps({})
for job in jobs:
set_default(self)
# limit local job to policy 0 for simplicty
if job['partition'] == '0' and job['policy_idx'] == 0:
local_job = job.copy()
continue
self.replicator.update(job)
self.assertEquals(mock_logger.exception.call_count, 0)
self.assertEquals(mock_logger.error.call_count, 0)
self.assertEquals(len(self.replicator.partition_times), 1)
self.assertEquals(self.replicator.suffix_hash, 0)
self.assertEquals(self.replicator.suffix_sync, 0)
self.assertEquals(self.replicator.suffix_count, 0)
mock_logger.reset_mock()
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
# Check successful http_connect and sync for local node
mock_tpool_reraise.return_value = (1, {'a83': 'ba47fd314242ec8c'
'7efb91f5d57336e4'})
resp.read.return_value = pickle.dumps({'a83': 'c130a2c17ed45102a'
'ada0f4eee69494ff'})
set_default(self)
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
self.replicator.sync = fake_func = mock.MagicMock()
self.replicator.update(local_job)
reqs = []
for node in local_job['nodes']:
reqs.append(mock.call(node, local_job, ['a83']))
fake_func.assert_has_calls(reqs, any_order=True)
self.assertEquals(fake_func.call_count, 2)
self.assertEquals(self.replicator.replication_count, 1)
self.assertEquals(self.replicator.suffix_sync, 2)
self.assertEquals(self.replicator.suffix_hash, 1)
self.assertEquals(self.replicator.suffix_count, 1)
mock_http.reset_mock()
mock_logger.reset_mock()
# test for replication params on policy 0 only
repl_job = local_job.copy()
for node in repl_job['nodes']:
node['replication_ip'] = '127.0.0.11'
node['replication_port'] = '6011'
set_default(self)
# with only one set of headers make sure we speicy index 0 here
# as otherwise it may be different from earlier tests
self.headers['X-Backend-Storage-Policy-Index'] = 0
self.replicator.update(repl_job)
reqs = []
for node in repl_job['nodes']:
reqs.append(mock.call(node['replication_ip'],
node['replication_port'], node['device'],
repl_job['partition'], 'REPLICATE',
'', headers=self.headers))
reqs.append(mock.call(node['replication_ip'],
node['replication_port'], node['device'],
repl_job['partition'], 'REPLICATE',
'/a83', headers=self.headers))
mock_http.assert_has_calls(reqs, any_order=True)
2010-07-12 17:03:45 -05:00
if __name__ == '__main__':
unittest.main()