swift/test/probe/test_object_async_update.py

233 lines
9.8 KiB
Python
Raw Normal View History

2010-07-12 17:03:45 -05:00
#!/usr/bin/python -u
# 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.
from io import StringIO
from unittest import main
2010-07-12 17:03:45 -05:00
from uuid import uuid4
from swiftclient import client
2010-07-12 17:03:45 -05:00
2016-05-17 14:22:05 +01:00
from swift.common import direct_client
from swift.common.manager import Manager
from test.probe.common import kill_nonprimary_server, \
2016-05-17 14:22:05 +01:00
kill_server, ReplProbeTest, start_server, ECProbeTest
2010-07-12 17:03:45 -05:00
class TestObjectAsyncUpdate(ReplProbeTest):
2010-07-12 17:03:45 -05:00
def test_main(self):
# Create container
2010-07-12 17:03:45 -05:00
container = 'container-%s' % uuid4()
client.put_container(self.url, self.token, container)
# Kill container servers excepting two of the primaries
2010-07-12 17:03:45 -05:00
cpart, cnodes = self.container_ring.get_nodes(self.account, container)
cnode = cnodes[0]
Allow 1+ object-servers-per-disk deployment Enabled by a new > 0 integer config value, "servers_per_port" in the [DEFAULT] config section for object-server and/or replication server configs. The setting's integer value determines how many different object-server workers handle requests for any single unique local port in the ring. In this mode, the parent swift-object-server process continues to run as the original user (i.e. root if low-port binding is required), binds to all ports as defined in the ring, and forks off the specified number of workers per listen socket. The child, per-port servers drop privileges and behave pretty much how object-server workers always have, except that because the ring has unique ports per disk, the object-servers will only be handling requests for a single disk. The parent process detects dead servers and restarts them (with the correct listen socket), starts missing servers when an updated ring file is found with a device on the server with a new port, and kills extraneous servers when their port is found to no longer be in the ring. The ring files are stat'ed at most every "ring_check_interval" seconds, as configured in the object-server config (same default of 15s). Immediately stopping all swift-object-worker processes still works by sending the parent a SIGTERM. Likewise, a SIGHUP to the parent process still causes the parent process to close all listen sockets and exit, allowing existing children to finish serving their existing requests. The drop_privileges helper function now has an optional param to suppress the setsid() call, which otherwise screws up the child workers' process management. The class method RingData.load() can be told to only load the ring metadata (i.e. everything except replica2part2dev_id) with the optional kwarg, header_only=True. This is used to keep the parent and all forked off workers from unnecessarily having full copies of all storage policy rings in memory. A new helper class, swift.common.storage_policy.BindPortsCache, provides a method to return a set of all device ports in all rings for the server on which it is instantiated (identified by its set of IP addresses). The BindPortsCache instance will track mtimes of ring files, so they are not opened more frequently than necessary. This patch includes enhancements to the probe tests and object-replicator/object-reconstructor config plumbing to allow the probe tests to work correctly both in the "normal" config (same IP but unique ports for each SAIO "server") and a server-per-port setup where each SAIO "server" must have a unique IP address and unique port per disk within each "server". The main probe tests only work with 4 servers and 4 disks, but you can see the difference in the rings for the EC probe tests where there are 2 disks per server for a total of 8 disks. Specifically, swift.common.ring.utils.is_local_device() will ignore the ports when the "my_port" argument is None. Then, object-replicator and object-reconstructor both set self.bind_port to None if server_per_port is enabled. Bonus improvement for IPv6 addresses in is_local_device(). This PR for vagrant-swift-all-in-one will aid in testing this patch: https://github.com/swiftstack/vagrant-swift-all-in-one/pull/16/ Also allow SAIO to answer is_local_device() better; common SAIO setups have multiple "servers" all on the same host with different ports for the different "servers" (which happen to match the IPs specified in the rings for the devices on each of those "servers"). However, you can configure the SAIO to have different localhost IP addresses (e.g. 127.0.0.1, 127.0.0.2, etc.) in the ring and in the servers' config files' bind_ip setting. This new whataremyips() implementation combined with a little plumbing allows is_local_device() to accurately answer, even on an SAIO. In the default case (an unspecified bind_ip defaults to '0.0.0.0') as well as an explict "bind to everything" like '0.0.0.0' or '::', whataremyips() behaves as it always has, returning all IP addresses for the server. Also updated probe tests to handle each "server" in the SAIO having a unique IP address. For some (noisy) benchmarks that show servers_per_port=X is at least as good as the same number of "normal" workers: https://gist.github.com/dbishop/c214f89ca708a6b1624a#file-summary-md Benchmarks showing the benefits of I/O isolation with a small number of slow disks: https://gist.github.com/dbishop/fd0ab067babdecfb07ca#file-results-md If you were wondering what the overhead of threads_per_disk looks like: https://gist.github.com/dbishop/1d14755fedc86a161718#file-tabular_results-md DocImpact Change-Id: I2239a4000b41a7e7cc53465ce794af49d44796c6
2015-05-14 22:14:15 -07:00
kill_nonprimary_server(cnodes, self.ipport2server, self.pids)
kill_server((cnode['ip'], cnode['port']),
self.ipport2server, self.pids)
# Create container/obj
2010-07-12 17:03:45 -05:00
obj = 'object-%s' % uuid4()
client.put_object(self.url, self.token, container, obj, '')
# Restart other primary server
Allow 1+ object-servers-per-disk deployment Enabled by a new > 0 integer config value, "servers_per_port" in the [DEFAULT] config section for object-server and/or replication server configs. The setting's integer value determines how many different object-server workers handle requests for any single unique local port in the ring. In this mode, the parent swift-object-server process continues to run as the original user (i.e. root if low-port binding is required), binds to all ports as defined in the ring, and forks off the specified number of workers per listen socket. The child, per-port servers drop privileges and behave pretty much how object-server workers always have, except that because the ring has unique ports per disk, the object-servers will only be handling requests for a single disk. The parent process detects dead servers and restarts them (with the correct listen socket), starts missing servers when an updated ring file is found with a device on the server with a new port, and kills extraneous servers when their port is found to no longer be in the ring. The ring files are stat'ed at most every "ring_check_interval" seconds, as configured in the object-server config (same default of 15s). Immediately stopping all swift-object-worker processes still works by sending the parent a SIGTERM. Likewise, a SIGHUP to the parent process still causes the parent process to close all listen sockets and exit, allowing existing children to finish serving their existing requests. The drop_privileges helper function now has an optional param to suppress the setsid() call, which otherwise screws up the child workers' process management. The class method RingData.load() can be told to only load the ring metadata (i.e. everything except replica2part2dev_id) with the optional kwarg, header_only=True. This is used to keep the parent and all forked off workers from unnecessarily having full copies of all storage policy rings in memory. A new helper class, swift.common.storage_policy.BindPortsCache, provides a method to return a set of all device ports in all rings for the server on which it is instantiated (identified by its set of IP addresses). The BindPortsCache instance will track mtimes of ring files, so they are not opened more frequently than necessary. This patch includes enhancements to the probe tests and object-replicator/object-reconstructor config plumbing to allow the probe tests to work correctly both in the "normal" config (same IP but unique ports for each SAIO "server") and a server-per-port setup where each SAIO "server" must have a unique IP address and unique port per disk within each "server". The main probe tests only work with 4 servers and 4 disks, but you can see the difference in the rings for the EC probe tests where there are 2 disks per server for a total of 8 disks. Specifically, swift.common.ring.utils.is_local_device() will ignore the ports when the "my_port" argument is None. Then, object-replicator and object-reconstructor both set self.bind_port to None if server_per_port is enabled. Bonus improvement for IPv6 addresses in is_local_device(). This PR for vagrant-swift-all-in-one will aid in testing this patch: https://github.com/swiftstack/vagrant-swift-all-in-one/pull/16/ Also allow SAIO to answer is_local_device() better; common SAIO setups have multiple "servers" all on the same host with different ports for the different "servers" (which happen to match the IPs specified in the rings for the devices on each of those "servers"). However, you can configure the SAIO to have different localhost IP addresses (e.g. 127.0.0.1, 127.0.0.2, etc.) in the ring and in the servers' config files' bind_ip setting. This new whataremyips() implementation combined with a little plumbing allows is_local_device() to accurately answer, even on an SAIO. In the default case (an unspecified bind_ip defaults to '0.0.0.0') as well as an explict "bind to everything" like '0.0.0.0' or '::', whataremyips() behaves as it always has, returning all IP addresses for the server. Also updated probe tests to handle each "server" in the SAIO having a unique IP address. For some (noisy) benchmarks that show servers_per_port=X is at least as good as the same number of "normal" workers: https://gist.github.com/dbishop/c214f89ca708a6b1624a#file-summary-md Benchmarks showing the benefits of I/O isolation with a small number of slow disks: https://gist.github.com/dbishop/fd0ab067babdecfb07ca#file-results-md If you were wondering what the overhead of threads_per_disk looks like: https://gist.github.com/dbishop/1d14755fedc86a161718#file-tabular_results-md DocImpact Change-Id: I2239a4000b41a7e7cc53465ce794af49d44796c6
2015-05-14 22:14:15 -07:00
start_server((cnode['ip'], cnode['port']),
self.ipport2server, self.pids)
# Assert it does not know about container/obj
self.assertFalse(direct_client.direct_get_container(
cnode, cpart, self.account, container)[1])
# Run the object-updaters
Manager(['object-updater']).once()
# Assert the other primary server now knows about container/obj
objs = [o['name'] for o in direct_client.direct_get_container(
cnode, cpart, self.account, container)[1]]
self.assertTrue(obj in objs)
2010-07-12 17:03:45 -05:00
class TestUpdateOverrides(ReplProbeTest):
"""
Use an internal client to PUT an object to proxy server,
bypassing gatekeeper so that X-Backend- headers can be included.
Verify that the update override headers take effect and override
values propagate to the container server.
"""
2016-05-17 14:22:05 +01:00
def test_update_during_PUT(self):
# verify that update sent during a PUT has override values
int_client = self.make_internal_client()
headers = {
'Content-Type': 'text/plain',
'X-Backend-Container-Update-Override-Etag': 'override-etag',
2016-05-17 14:22:05 +01:00
'X-Backend-Container-Update-Override-Content-Type':
'override-type',
'X-Backend-Container-Update-Override-Size': '1999'
}
client.put_container(self.url, self.token, 'c1',
headers={'X-Storage-Policy':
self.policy.name})
2016-05-17 14:22:05 +01:00
int_client.upload_object(
StringIO(u'stuff'), self.account, 'c1', 'o1', headers)
# Run the object-updaters to be sure updates are done
Manager(['object-updater']).once()
2016-05-17 14:22:05 +01:00
meta = int_client.get_object_metadata(self.account, 'c1', 'o1')
self.assertEqual('text/plain', meta['content-type'])
self.assertEqual('c13d88cb4cb02003daedb8a84e5d272a', meta['etag'])
2016-05-17 14:22:05 +01:00
self.assertEqual('5', meta['content-length'])
2016-05-17 14:22:05 +01:00
obj_iter = int_client.iter_objects(self.account, 'c1')
for obj in obj_iter:
if obj['name'] == 'o1':
self.assertEqual('override-etag', obj['hash'])
self.assertEqual('override-type', obj['content_type'])
2016-05-17 14:22:05 +01:00
self.assertEqual(1999, obj['bytes'])
break
else:
self.fail('Failed to find object o1 in listing')
2016-05-17 14:22:05 +01:00
class TestUpdateOverridesEC(ECProbeTest):
# verify that the container update overrides used with EC policies make
# it to the container servers when container updates are sync or async
# and possibly re-ordered with respect to object PUT and POST requests.
def test_async_update_after_PUT(self):
cpart, cnodes = self.container_ring.get_nodes(self.account, 'c1')
client.put_container(self.url, self.token, 'c1',
headers={'X-Storage-Policy':
self.policy.name})
# put an object while one container server is stopped so that we force
# an async update to it
kill_server((cnodes[0]['ip'], cnodes[0]['port']),
self.ipport2server, self.pids)
content = u'stuff'
client.put_object(self.url, self.token, 'c1', 'o1', contents=content)
meta = client.head_object(self.url, self.token, 'c1', 'o1')
# re-start the container server and assert that it does not yet know
# about the object
start_server((cnodes[0]['ip'], cnodes[0]['port']),
self.ipport2server, self.pids)
self.assertFalse(direct_client.direct_get_container(
cnodes[0], cpart, self.account, 'c1')[1])
# Run the object-updaters to be sure updates are done
Manager(['object-updater']).once()
# check the re-started container server has update with override values
obj = direct_client.direct_get_container(
cnodes[0], cpart, self.account, 'c1')[1][0]
self.assertEqual(meta['etag'], obj['hash'])
self.assertEqual(len(content), obj['bytes'])
def test_update_during_POST_only(self):
# verify correct update values when PUT update is missed but then a
# POST update succeeds *before* the PUT async pending update is sent
cpart, cnodes = self.container_ring.get_nodes(self.account, 'c1')
client.put_container(self.url, self.token, 'c1',
headers={'X-Storage-Policy':
self.policy.name})
# put an object while one container server is stopped so that we force
# an async update to it
kill_server((cnodes[0]['ip'], cnodes[0]['port']),
self.ipport2server, self.pids)
content = u'stuff'
client.put_object(self.url, self.token, 'c1', 'o1', contents=content)
meta = client.head_object(self.url, self.token, 'c1', 'o1')
# re-start the container server and assert that it does not yet know
# about the object
start_server((cnodes[0]['ip'], cnodes[0]['port']),
self.ipport2server, self.pids)
self.assertFalse(direct_client.direct_get_container(
cnodes[0], cpart, self.account, 'c1')[1])
# use internal client for POST so we can force fast-post mode
int_client = self.make_internal_client(object_post_as_copy=False)
int_client.set_object_metadata(
self.account, 'c1', 'o1', {'X-Object-Meta-Fruit': 'Tomato'})
self.assertEqual(
'Tomato',
int_client.get_object_metadata(self.account, 'c1', 'o1')
['x-object-meta-fruit']) # sanity
# check the re-started container server has update with override values
obj = direct_client.direct_get_container(
cnodes[0], cpart, self.account, 'c1')[1][0]
self.assertEqual(meta['etag'], obj['hash'])
self.assertEqual(len(content), obj['bytes'])
# Run the object-updaters to send the async pending from the PUT
Manager(['object-updater']).once()
# check container listing metadata is still correct
obj = direct_client.direct_get_container(
cnodes[0], cpart, self.account, 'c1')[1][0]
self.assertEqual(meta['etag'], obj['hash'])
self.assertEqual(len(content), obj['bytes'])
def test_async_updates_after_PUT_and_POST(self):
# verify correct update values when PUT update and POST updates are
# missed but then async updates are sent
cpart, cnodes = self.container_ring.get_nodes(self.account, 'c1')
client.put_container(self.url, self.token, 'c1',
headers={'X-Storage-Policy':
self.policy.name})
# PUT and POST to object while one container server is stopped so that
# we force async updates to it
kill_server((cnodes[0]['ip'], cnodes[0]['port']),
self.ipport2server, self.pids)
content = u'stuff'
client.put_object(self.url, self.token, 'c1', 'o1', contents=content)
meta = client.head_object(self.url, self.token, 'c1', 'o1')
# use internal client for POST so we can force fast-post mode
int_client = self.make_internal_client(object_post_as_copy=False)
int_client.set_object_metadata(
self.account, 'c1', 'o1', {'X-Object-Meta-Fruit': 'Tomato'})
self.assertEqual(
'Tomato',
int_client.get_object_metadata(self.account, 'c1', 'o1')
['x-object-meta-fruit']) # sanity
# re-start the container server and assert that it does not yet know
# about the object
start_server((cnodes[0]['ip'], cnodes[0]['port']),
self.ipport2server, self.pids)
self.assertFalse(direct_client.direct_get_container(
cnodes[0], cpart, self.account, 'c1')[1])
# Run the object-updaters to send the async pendings
Manager(['object-updater']).once()
# check container listing metadata is still correct
obj = direct_client.direct_get_container(
cnodes[0], cpart, self.account, 'c1')[1][0]
self.assertEqual(meta['etag'], obj['hash'])
self.assertEqual(len(content), obj['bytes'])
2010-07-12 17:03:45 -05:00
if __name__ == '__main__':
main()