1819 lines
78 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.
"""Tests for swift.common.db"""
import contextlib
2010-07-12 17:03:45 -05:00
import os
import sys
2010-07-12 17:03:45 -05:00
import unittest
from tempfile import mkdtemp
2011-08-02 18:21:25 +00:00
from shutil import rmtree, copy
2010-07-12 17:03:45 -05:00
from uuid import uuid4
cli: add --sync to db info to show syncs When looking at containers and accounts it's sometimes nice to know who they've been replicating with. This patch adds a `--sync|-s` option to swift-{container|account}-info which will also dump the incoming and outgoing sync tables: $ swift-container-info /srv/node3/sdb3/containers/294/624/49b9ff074c502ec5e429e7af99a30624/49b9ff074c502ec5e429e7af99a30624.db -s Path: /AUTH_test/new Account: AUTH_test Container: new Deleted: False Container Hash: 49b9ff074c502ec5e429e7af99a30624 Metadata: Created at: 2022-02-16T05:34:05.988480 (1644989645.98848) Put Timestamp: 2022-02-16T05:34:05.981320 (1644989645.98132) Delete Timestamp: 1970-01-01T00:00:00.000000 (0) Status Timestamp: 2022-02-16T05:34:05.981320 (1644989645.98132) Object Count: 1 Bytes Used: 7 Storage Policy: default (0) Reported Put Timestamp: 1970-01-01T00:00:00.000000 (0) Reported Delete Timestamp: 1970-01-01T00:00:00.000000 (0) Reported Object Count: 0 Reported Bytes Used: 0 Chexor: 962368324c2ca023c56669d03ed92807 UUID: f33184e7-56d5-4c74-9d2e-5417c187d722-sdb3 X-Container-Sync-Point2: -1 X-Container-Sync-Point1: -1 No system metadata found in db file No user metadata found in db file Sharding Metadata: Type: root State: unsharded Incoming Syncs: Sync Point Remote ID Updated At 1 ce7268a1-f5d0-4b83-b993-af17b602a0ff-sdb1 2022-02-16T05:38:22.000000 (1644989902) 1 2af5abc0-7f70-4e2f-8f94-737aeaada7f4-sdb4 2022-02-16T05:38:22.000000 (1644989902) Outgoing Syncs: Sync Point Remote ID Updated At Partition 294 Hash 49b9ff074c502ec5e429e7af99a30624 As a follow up to the device in DB ID patch we can see that the replicas at sdb1 and sdb4 have replicated with this node. Change-Id: I23d786e82c6710bea7660a9acf8bbbd113b5b727
2022-02-16 16:38:52 +11:00
import mock
import six.moves.cPickle as pickle
2010-07-12 17:03:45 -05:00
import base64
import json
2010-07-12 17:03:45 -05:00
import sqlite3
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
import itertools
import time
import random
from mock import patch, MagicMock
from eventlet.timeout import Timeout
from six.moves import range
2010-07-12 17:03:45 -05:00
import six
2011-08-01 20:46:30 +00:00
import swift.common.db
from swift.common.constraints import \
MAX_META_VALUE_LENGTH, MAX_META_COUNT, MAX_META_OVERALL_SIZE
from swift.common.db import chexor, dict_factory, get_db_connection, \
DatabaseBroker, DatabaseConnectionError, DatabaseAlreadyExists, \
GreenDBConnection, PICKLE_PROTOCOL, zero_like, TombstoneReclaimer
from swift.common.utils import normalize_timestamp, mkdirs, Timestamp
2010-07-12 17:03:45 -05:00
from swift.common.exceptions import LockTimeout
from swift.common.swob import HTTPException
2010-07-12 17:03:45 -05:00
from test.unit import make_timestamp_iter, generate_db_path
2010-07-12 17:03:45 -05:00
class TestHelperFunctions(unittest.TestCase):
def test_zero_like(self):
expectations = {
# value => expected
None: True,
True: False,
'': True,
'asdf': False,
0: True,
1: False,
'0': True,
'1': False,
}
errors = []
for value, expected in expectations.items():
rv = zero_like(value)
if rv != expected:
errors.append('zero_like(%r) => %r expected %r' % (
value, rv, expected))
if errors:
self.fail('Some unexpected return values:\n' + '\n'.join(errors))
2010-07-12 17:03:45 -05:00
class TestDatabaseConnectionError(unittest.TestCase):
def test_str(self):
err = \
DatabaseConnectionError(':memory:', 'No valid database connection')
self.assertIn(':memory:', str(err))
self.assertIn('No valid database connection', str(err))
2010-07-12 17:03:45 -05:00
err = DatabaseConnectionError(':memory:',
'No valid database connection',
timeout=1357)
self.assertIn(':memory:', str(err))
self.assertIn('No valid database connection', str(err))
self.assertIn('1357', str(err))
2010-07-12 17:03:45 -05:00
class TestDictFactory(unittest.TestCase):
def test_normal_case(self):
conn = sqlite3.connect(':memory:')
conn.execute('CREATE TABLE test (one TEXT, two INTEGER)')
conn.execute('INSERT INTO test (one, two) VALUES ("abc", 123)')
conn.execute('INSERT INTO test (one, two) VALUES ("def", 456)')
conn.commit()
curs = conn.execute('SELECT one, two FROM test')
self.assertEqual(dict_factory(curs, next(curs)),
{'one': 'abc', 'two': 123})
self.assertEqual(dict_factory(curs, next(curs)),
{'one': 'def', 'two': 456})
2010-07-12 17:03:45 -05:00
class TestChexor(unittest.TestCase):
def test_normal_case(self):
self.assertEqual(
chexor('d41d8cd98f00b204e9800998ecf8427e',
'new name', normalize_timestamp(1)),
2010-07-12 17:03:45 -05:00
'4f2ea31ac14d4273fe32ba08062b21de')
def test_invalid_old_hash(self):
self.assertRaises(ValueError, chexor, 'oldhash', 'name',
2010-07-12 17:03:45 -05:00
normalize_timestamp(1))
def test_no_name(self):
self.assertRaises(Exception, chexor,
'd41d8cd98f00b204e9800998ecf8427e', None,
normalize_timestamp(1))
2010-07-12 17:03:45 -05:00
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
def test_chexor(self):
ts = (normalize_timestamp(ts) for ts in
itertools.count(int(time.time())))
objects = [
('frank', next(ts)),
('bob', next(ts)),
('tom', next(ts)),
('frank', next(ts)),
('tom', next(ts)),
('bob', next(ts)),
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
]
hash_ = '0'
random.shuffle(objects)
for obj in objects:
hash_ = chexor(hash_, *obj)
other_hash = '0'
random.shuffle(objects)
for obj in objects:
other_hash = chexor(other_hash, *obj)
self.assertEqual(hash_, other_hash)
2010-07-12 17:03:45 -05:00
class TestGreenDBConnection(unittest.TestCase):
def test_execute_when_locked(self):
# This test is dependent on the code under test calling execute and
# commit as sqlite3.Cursor.execute in a subclass.
class InterceptCursor(sqlite3.Cursor):
pass
db_error = sqlite3.OperationalError('database is locked')
InterceptCursor.execute = MagicMock(side_effect=db_error)
with patch('sqlite3.Cursor', new=InterceptCursor):
conn = sqlite3.connect(':memory:', check_same_thread=False,
factory=GreenDBConnection, timeout=0.1)
self.assertRaises(Timeout, conn.execute, 'select 1')
self.assertTrue(InterceptCursor.execute.called)
self.assertEqual(InterceptCursor.execute.call_args_list,
list((InterceptCursor.execute.call_args,) *
InterceptCursor.execute.call_count))
def text_commit_when_locked(self):
# This test is dependent on the code under test calling commit and
# commit as sqlite3.Connection.commit in a subclass.
class InterceptConnection(sqlite3.Connection):
pass
db_error = sqlite3.OperationalError('database is locked')
InterceptConnection.commit = MagicMock(side_effect=db_error)
with patch('sqlite3.Connection', new=InterceptConnection):
conn = sqlite3.connect(':memory:', check_same_thread=False,
factory=GreenDBConnection, timeout=0.1)
self.assertRaises(Timeout, conn.commit)
self.assertTrue(InterceptConnection.commit.called)
self.assertEqual(InterceptConnection.commit.call_args_list,
list((InterceptConnection.commit.call_args,) *
InterceptConnection.commit.call_count))
class TestDbBase(unittest.TestCase):
server_type = 'container'
testdir = None
def setUp(self):
self.testdir = mkdtemp()
self.db_path = self.get_db_path()
def tearDown(self):
rmtree(self.testdir, ignore_errors=True)
def get_db_path(self):
return generate_db_path(self.testdir, self.server_type)
class TestGetDBConnection(TestDbBase):
def setUp(self):
super(TestGetDBConnection, self).setUp()
self.db_path = self.init_db_path()
def init_db_path(self):
# Test ContainerBroker.empty
db_path = self.get_db_path()
broker = ExampleBroker(db_path, account='a')
broker.initialize(Timestamp.now().internal, 0)
return db_path
2010-07-12 17:03:45 -05:00
def test_normal_case(self):
conn = get_db_connection(self.db_path)
self.assertTrue(hasattr(conn, 'execute'))
2010-07-12 17:03:45 -05:00
def test_invalid_path(self):
self.assertRaises(DatabaseConnectionError, get_db_connection,
'invalid database path / name')
def test_locked_db(self):
# This test is dependent on the code under test calling execute and
# commit as sqlite3.Cursor.execute in a subclass.
class InterceptCursor(sqlite3.Cursor):
pass
db_error = sqlite3.OperationalError('database is locked')
mock_db_cmd = MagicMock(side_effect=db_error)
InterceptCursor.execute = mock_db_cmd
with patch('sqlite3.Cursor', new=InterceptCursor):
self.assertRaises(Timeout, get_db_connection,
self.db_path, timeout=0.1)
self.assertTrue(mock_db_cmd.called)
self.assertEqual(mock_db_cmd.call_args_list,
list((mock_db_cmd.call_args,) *
mock_db_cmd.call_count))
2010-07-12 17:03:45 -05:00
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
class ExampleBroker(DatabaseBroker):
"""
Concrete enough implementation of a DatabaseBroker.
"""
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
db_type = 'test'
db_contains_type = 'test'
db_reclaim_timestamp = 'created_at'
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
def _initialize(self, conn, put_timestamp, **kwargs):
if not self.account:
raise ValueError(
'Attempting to create a new database with no account set')
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
conn.executescript('''
CREATE TABLE test_stat (
account TEXT,
test_count INTEGER DEFAULT 0,
created_at TEXT,
put_timestamp TEXT DEFAULT '0',
delete_timestamp TEXT DEFAULT '0',
hash TEXT default '00000000000000000000000000000000',
id TEXT,
status TEXT DEFAULT '',
status_changed_at TEXT DEFAULT '0',
metadata TEXT DEFAULT ''
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
);
CREATE TABLE test (
ROWID INTEGER PRIMARY KEY AUTOINCREMENT,
name TEXT,
created_at TEXT,
deleted INTEGER DEFAULT 0
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
);
CREATE TRIGGER test_insert AFTER INSERT ON test
BEGIN
UPDATE test_stat
SET test_count = test_count + (1 - new.deleted);
END;
CREATE TRIGGER test_delete AFTER DELETE ON test
BEGIN
UPDATE test_stat
SET test_count = test_count - (1 - old.deleted);
END;
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
''')
conn.execute("""
INSERT INTO test_stat (
account, created_at, id, put_timestamp, status_changed_at, status)
VALUES (?, ?, ?, ?, ?, ?);
""", (self.account, Timestamp.now().internal, str(uuid4()),
put_timestamp, put_timestamp, ''))
def merge_items(self, item_list):
with self.get() as conn:
for rec in item_list:
conn.execute(
'DELETE FROM test WHERE name = ? and created_at < ?', (
rec['name'], rec['created_at']))
if not conn.execute(
'SELECT 1 FROM test WHERE name = ?',
(rec['name'],)).fetchall():
conn.execute('''
INSERT INTO test (name, created_at, deleted)
VALUES (?, ?, ?)''', (
rec['name'], rec['created_at'], rec['deleted']))
conn.commit()
def _commit_puts_load(self, item_list, entry):
(name, timestamp, deleted) = entry
item_list.append({
'name': name,
'created_at': timestamp,
'deleted': deleted,
})
def _load_item(self, name, timestamp, deleted):
if self.db_file == ':memory:':
record = {
'name': name,
'created_at': timestamp,
'deleted': deleted,
}
self.merge_items([record])
return
with open(self.pending_file, 'a+b') as fp:
fp.write(b':')
fp.write(base64.b64encode(pickle.dumps(
(name, timestamp, deleted),
protocol=PICKLE_PROTOCOL)))
fp.flush()
def put_test(self, name, timestamp):
self._load_item(name, timestamp, 0)
def delete_test(self, name, timestamp):
self._load_item(name, timestamp, 1)
def _delete_db(self, conn, timestamp):
conn.execute("""
UPDATE test_stat
SET delete_timestamp = ?,
status = 'DELETED',
status_changed_at = ?
WHERE delete_timestamp < ? """, (timestamp, timestamp, timestamp))
def _is_deleted(self, conn):
info = conn.execute('SELECT * FROM test_stat').fetchone()
return (info['test_count'] in (None, '', 0, '0')) and \
Add two vector timestamps The normalized form of the X-Timestamp header looks like a float with a fixed width to ensure stable string sorting - normalized timestamps look like "1402464677.04188" To support overwrites of existing data without modifying the original timestamp but still maintain consistency a second internal offset vector is append to the normalized timestamp form which compares and sorts greater than the fixed width float format but less than a newer timestamp. The internalized format of timestamps looks like "1402464677.04188_0000000000000000" - the portion after the underscore is the offset and is a formatted hexadecimal integer. The internalized form is not exposed to clients in responses from Swift. Normal client operations will not create a timestamp with an offset. The Timestamp class in common.utils supports internalized and normalized formatting of timestamps and also comparison of timestamp values. When the offset value of a Timestamp is 0 - it's considered insignificant and need not be represented in the string format; to support backwards compatibility during a Swift upgrade the internalized and normalized form of a Timestamp with an insignificant offset are identical. When a timestamp includes an offset it will always be represented in the internalized form, but is still excluded from the normalized form. Timestamps with an equivalent timestamp portion (the float part) will compare and order by their offset. Timestamps with a greater timestamp portion will always compare and order greater than a Timestamp with a lesser timestamp regardless of it's offset. String comparison and ordering is guaranteed for the internalized string format, and is backwards compatible for normalized timestamps which do not include an offset. The reconciler currently uses a offset bump to ensure that objects can move to the wrong storage policy and be moved back. This use-case is valid because the content represented by the user-facing timestamp is not modified in way. Future consumers of the offset vector of timestamps should be mindful of HTTP semantics of If-Modified and take care to avoid deviation in the response from the object server without an accompanying change to the user facing timestamp. DocImpact Implements: blueprint storage-policies Change-Id: Id85c960b126ec919a481dc62469bf172b7fb8549
2014-06-10 22:17:47 -07:00
(Timestamp(info['delete_timestamp']) >
Timestamp(info['put_timestamp']))
class TestExampleBroker(TestDbBase):
"""
Tests that use the mostly Concrete enough ExampleBroker to exercise some
of the abstract methods on DatabaseBroker.
"""
broker_class = ExampleBroker
policy = 0
server_type = 'example'
def setUp(self):
super(TestExampleBroker, self).setUp()
self.ts = make_timestamp_iter()
def test_delete_db(self):
broker = self.broker_class(self.db_path, account='a', container='c')
broker.initialize(next(self.ts).internal)
broker.delete_db(next(self.ts).internal)
self.assertTrue(broker.is_deleted())
def test_merge_timestamps_simple_delete(self):
put_timestamp = next(self.ts).internal
broker = self.broker_class(self.db_path, account='a', container='c')
broker.initialize(put_timestamp)
created_at = broker.get_info()['created_at']
broker.merge_timestamps(created_at, put_timestamp, '0')
info = broker.get_info()
self.assertEqual(info['created_at'], created_at)
self.assertEqual(info['put_timestamp'], put_timestamp)
self.assertEqual(info['delete_timestamp'], '0')
self.assertEqual(info['status_changed_at'], put_timestamp)
# delete
delete_timestamp = next(self.ts).internal
broker.merge_timestamps(created_at, put_timestamp, delete_timestamp)
self.assertTrue(broker.is_deleted())
info = broker.get_info()
self.assertEqual(info['created_at'], created_at)
self.assertEqual(info['put_timestamp'], put_timestamp)
self.assertEqual(info['delete_timestamp'], delete_timestamp)
self.assertTrue(info['status_changed_at'] > Timestamp(put_timestamp))
def put_item(self, broker, timestamp):
broker.put_test('test', timestamp)
def delete_item(self, broker, timestamp):
broker.delete_test('test', timestamp)
def test_merge_timestamps_delete_with_objects(self):
put_timestamp = next(self.ts).internal
broker = self.broker_class(self.db_path, account='a', container='c')
broker.initialize(put_timestamp, storage_policy_index=int(self.policy))
created_at = broker.get_info()['created_at']
broker.merge_timestamps(created_at, put_timestamp, '0')
info = broker.get_info()
self.assertEqual(info['created_at'], created_at)
self.assertEqual(info['put_timestamp'], put_timestamp)
self.assertEqual(info['delete_timestamp'], '0')
self.assertEqual(info['status_changed_at'], put_timestamp)
# add object
self.put_item(broker, next(self.ts).internal)
self.assertEqual(broker.get_info()[
'%s_count' % broker.db_contains_type], 1)
# delete
delete_timestamp = next(self.ts).internal
broker.merge_timestamps(created_at, put_timestamp, delete_timestamp)
self.assertFalse(broker.is_deleted())
info = broker.get_info()
self.assertEqual(info['created_at'], created_at)
self.assertEqual(info['put_timestamp'], put_timestamp)
self.assertEqual(info['delete_timestamp'], delete_timestamp)
# status is unchanged
self.assertEqual(info['status_changed_at'], put_timestamp)
# count is causing status to hold on
self.delete_item(broker, next(self.ts).internal)
self.assertEqual(broker.get_info()[
'%s_count' % broker.db_contains_type], 0)
self.assertTrue(broker.is_deleted())
def test_merge_timestamps_simple_recreate(self):
put_timestamp = next(self.ts).internal
broker = self.broker_class(self.db_path, account='a', container='c')
broker.initialize(put_timestamp, storage_policy_index=int(self.policy))
virgin_status_changed_at = broker.get_info()['status_changed_at']
created_at = broker.get_info()['created_at']
delete_timestamp = next(self.ts).internal
broker.merge_timestamps(created_at, put_timestamp, delete_timestamp)
self.assertTrue(broker.is_deleted())
info = broker.get_info()
self.assertEqual(info['created_at'], created_at)
self.assertEqual(info['put_timestamp'], put_timestamp)
self.assertEqual(info['delete_timestamp'], delete_timestamp)
orig_status_changed_at = info['status_changed_at']
self.assertTrue(orig_status_changed_at >
Timestamp(virgin_status_changed_at))
# recreate
recreate_timestamp = next(self.ts).internal
status_changed_at = time.time()
with patch('swift.common.db.time.time', new=lambda: status_changed_at):
broker.merge_timestamps(created_at, recreate_timestamp, '0')
self.assertFalse(broker.is_deleted())
info = broker.get_info()
self.assertEqual(info['created_at'], created_at)
self.assertEqual(info['put_timestamp'], recreate_timestamp)
self.assertEqual(info['delete_timestamp'], delete_timestamp)
self.assertEqual(Timestamp(status_changed_at).normal,
info['status_changed_at'])
def test_merge_timestamps_recreate_with_objects(self):
put_timestamp = next(self.ts).internal
broker = self.broker_class(self.db_path, account='a', container='c')
broker.initialize(put_timestamp, storage_policy_index=int(self.policy))
created_at = broker.get_info()['created_at']
# delete
delete_timestamp = next(self.ts).internal
broker.merge_timestamps(created_at, put_timestamp, delete_timestamp)
self.assertTrue(broker.is_deleted())
info = broker.get_info()
self.assertEqual(info['created_at'], created_at)
self.assertEqual(info['put_timestamp'], put_timestamp)
self.assertEqual(info['delete_timestamp'], delete_timestamp)
orig_status_changed_at = info['status_changed_at']
self.assertTrue(Timestamp(orig_status_changed_at) >=
Timestamp(put_timestamp))
# add object
self.put_item(broker, next(self.ts).internal)
count_key = '%s_count' % broker.db_contains_type
self.assertEqual(broker.get_info()[count_key], 1)
self.assertFalse(broker.is_deleted())
# recreate
recreate_timestamp = next(self.ts).internal
broker.merge_timestamps(created_at, recreate_timestamp, '0')
self.assertFalse(broker.is_deleted())
info = broker.get_info()
self.assertEqual(info['created_at'], created_at)
self.assertEqual(info['put_timestamp'], recreate_timestamp)
self.assertEqual(info['delete_timestamp'], delete_timestamp)
self.assertEqual(info['status_changed_at'], orig_status_changed_at)
# count is not causing status to hold on
self.delete_item(broker, next(self.ts).internal)
self.assertFalse(broker.is_deleted())
def test_merge_timestamps_update_put_no_status_change(self):
put_timestamp = next(self.ts).internal
broker = self.broker_class(self.db_path, account='a', container='c')
broker.initialize(put_timestamp, storage_policy_index=int(self.policy))
info = broker.get_info()
orig_status_changed_at = info['status_changed_at']
created_at = info['created_at']
new_put_timestamp = next(self.ts).internal
broker.merge_timestamps(created_at, new_put_timestamp, '0')
info = broker.get_info()
self.assertEqual(new_put_timestamp, info['put_timestamp'])
self.assertEqual(orig_status_changed_at, info['status_changed_at'])
def test_merge_timestamps_update_delete_no_status_change(self):
put_timestamp = next(self.ts).internal
broker = self.broker_class(self.db_path, account='a', container='c')
broker.initialize(put_timestamp, storage_policy_index=int(self.policy))
created_at = broker.get_info()['created_at']
broker.merge_timestamps(created_at, put_timestamp,
next(self.ts).internal)
orig_status_changed_at = broker.get_info()['status_changed_at']
new_delete_timestamp = next(self.ts).internal
broker.merge_timestamps(created_at, put_timestamp,
new_delete_timestamp)
info = broker.get_info()
self.assertEqual(new_delete_timestamp, info['delete_timestamp'])
self.assertEqual(orig_status_changed_at, info['status_changed_at'])
def test_get_max_row(self):
broker = self.broker_class(self.db_path, account='a', container='c')
broker.initialize(next(self.ts).internal,
storage_policy_index=int(self.policy))
self.assertEqual(-1, broker.get_max_row())
self.put_item(broker, next(self.ts).internal)
# commit pending file into db
broker._commit_puts()
self.assertEqual(1, broker.get_max_row())
self.delete_item(broker, next(self.ts).internal)
broker._commit_puts()
self.assertEqual(2, broker.get_max_row())
self.put_item(broker, next(self.ts).internal)
broker._commit_puts()
self.assertEqual(3, broker.get_max_row())
def test_get_info(self):
broker = self.broker_class(self.db_path, account='test', container='c')
created_at = time.time()
with patch('swift.common.db.time.time', new=lambda: created_at):
Add two vector timestamps The normalized form of the X-Timestamp header looks like a float with a fixed width to ensure stable string sorting - normalized timestamps look like "1402464677.04188" To support overwrites of existing data without modifying the original timestamp but still maintain consistency a second internal offset vector is append to the normalized timestamp form which compares and sorts greater than the fixed width float format but less than a newer timestamp. The internalized format of timestamps looks like "1402464677.04188_0000000000000000" - the portion after the underscore is the offset and is a formatted hexadecimal integer. The internalized form is not exposed to clients in responses from Swift. Normal client operations will not create a timestamp with an offset. The Timestamp class in common.utils supports internalized and normalized formatting of timestamps and also comparison of timestamp values. When the offset value of a Timestamp is 0 - it's considered insignificant and need not be represented in the string format; to support backwards compatibility during a Swift upgrade the internalized and normalized form of a Timestamp with an insignificant offset are identical. When a timestamp includes an offset it will always be represented in the internalized form, but is still excluded from the normalized form. Timestamps with an equivalent timestamp portion (the float part) will compare and order by their offset. Timestamps with a greater timestamp portion will always compare and order greater than a Timestamp with a lesser timestamp regardless of it's offset. String comparison and ordering is guaranteed for the internalized string format, and is backwards compatible for normalized timestamps which do not include an offset. The reconciler currently uses a offset bump to ensure that objects can move to the wrong storage policy and be moved back. This use-case is valid because the content represented by the user-facing timestamp is not modified in way. Future consumers of the offset vector of timestamps should be mindful of HTTP semantics of If-Modified and take care to avoid deviation in the response from the object server without an accompanying change to the user facing timestamp. DocImpact Implements: blueprint storage-policies Change-Id: Id85c960b126ec919a481dc62469bf172b7fb8549
2014-06-10 22:17:47 -07:00
broker.initialize(Timestamp(1).internal,
storage_policy_index=int(self.policy))
info = broker.get_info()
count_key = '%s_count' % broker.db_contains_type
expected = {
count_key: 0,
Add two vector timestamps The normalized form of the X-Timestamp header looks like a float with a fixed width to ensure stable string sorting - normalized timestamps look like "1402464677.04188" To support overwrites of existing data without modifying the original timestamp but still maintain consistency a second internal offset vector is append to the normalized timestamp form which compares and sorts greater than the fixed width float format but less than a newer timestamp. The internalized format of timestamps looks like "1402464677.04188_0000000000000000" - the portion after the underscore is the offset and is a formatted hexadecimal integer. The internalized form is not exposed to clients in responses from Swift. Normal client operations will not create a timestamp with an offset. The Timestamp class in common.utils supports internalized and normalized formatting of timestamps and also comparison of timestamp values. When the offset value of a Timestamp is 0 - it's considered insignificant and need not be represented in the string format; to support backwards compatibility during a Swift upgrade the internalized and normalized form of a Timestamp with an insignificant offset are identical. When a timestamp includes an offset it will always be represented in the internalized form, but is still excluded from the normalized form. Timestamps with an equivalent timestamp portion (the float part) will compare and order by their offset. Timestamps with a greater timestamp portion will always compare and order greater than a Timestamp with a lesser timestamp regardless of it's offset. String comparison and ordering is guaranteed for the internalized string format, and is backwards compatible for normalized timestamps which do not include an offset. The reconciler currently uses a offset bump to ensure that objects can move to the wrong storage policy and be moved back. This use-case is valid because the content represented by the user-facing timestamp is not modified in way. Future consumers of the offset vector of timestamps should be mindful of HTTP semantics of If-Modified and take care to avoid deviation in the response from the object server without an accompanying change to the user facing timestamp. DocImpact Implements: blueprint storage-policies Change-Id: Id85c960b126ec919a481dc62469bf172b7fb8549
2014-06-10 22:17:47 -07:00
'created_at': Timestamp(created_at).internal,
'put_timestamp': Timestamp(1).internal,
'status_changed_at': Timestamp(1).internal,
'delete_timestamp': '0',
}
for k, v in expected.items():
self.assertEqual(info[k], v,
'mismatch for %s, %s != %s' % (
k, info[k], v))
def test_get_raw_metadata(self):
broker = self.broker_class(self.db_path, account='test', container='c')
Add two vector timestamps The normalized form of the X-Timestamp header looks like a float with a fixed width to ensure stable string sorting - normalized timestamps look like "1402464677.04188" To support overwrites of existing data without modifying the original timestamp but still maintain consistency a second internal offset vector is append to the normalized timestamp form which compares and sorts greater than the fixed width float format but less than a newer timestamp. The internalized format of timestamps looks like "1402464677.04188_0000000000000000" - the portion after the underscore is the offset and is a formatted hexadecimal integer. The internalized form is not exposed to clients in responses from Swift. Normal client operations will not create a timestamp with an offset. The Timestamp class in common.utils supports internalized and normalized formatting of timestamps and also comparison of timestamp values. When the offset value of a Timestamp is 0 - it's considered insignificant and need not be represented in the string format; to support backwards compatibility during a Swift upgrade the internalized and normalized form of a Timestamp with an insignificant offset are identical. When a timestamp includes an offset it will always be represented in the internalized form, but is still excluded from the normalized form. Timestamps with an equivalent timestamp portion (the float part) will compare and order by their offset. Timestamps with a greater timestamp portion will always compare and order greater than a Timestamp with a lesser timestamp regardless of it's offset. String comparison and ordering is guaranteed for the internalized string format, and is backwards compatible for normalized timestamps which do not include an offset. The reconciler currently uses a offset bump to ensure that objects can move to the wrong storage policy and be moved back. This use-case is valid because the content represented by the user-facing timestamp is not modified in way. Future consumers of the offset vector of timestamps should be mindful of HTTP semantics of If-Modified and take care to avoid deviation in the response from the object server without an accompanying change to the user facing timestamp. DocImpact Implements: blueprint storage-policies Change-Id: Id85c960b126ec919a481dc62469bf172b7fb8549
2014-06-10 22:17:47 -07:00
broker.initialize(Timestamp(0).internal,
storage_policy_index=int(self.policy))
self.assertEqual(broker.metadata, {})
self.assertEqual(broker.get_raw_metadata(), '')
# This is not obvious. The actual JSON in the database is the same:
# '{"test\\u062a": ["value\\u062a", "0000000001.00000"]}'
# The only difference is what reading it produces on py2 and py3.
sharding: Better-handle newlines in container names Previously, if you were on Python 2.7.10+ [0], such a newline would cause the sharder to fail, complaining about invalid header values when trying to create the shard containers. On older versions of Python, it would most likely cause a parsing error in the container-server that was trying to handle the PUT. Now, quote all places that we pass around container paths. This includes: * The X-Container-Sysmeta-Shard-(Quoted-)Root sent when creating the (empty) remote shards * The X-Container-Sysmeta-Shard-(Quoted-)Root included when initializing the local handoff for cleaving * The X-Backend-(Quoted-)Container-Path the proxy sends to the object-server for container updates * The Location header the container-server sends to the object-updater Note that a new header was required in requests so that servers would know whether the value should be unquoted or not. We can get away with reusing Location in responses by having clients opt-in to quoting with a new X-Backend-Accept-Quoted-Location header. During a rolling upgrade, * old object-servers servicing requests from new proxy-servers will not know about the container path override and so will try to update the root container, * in general, object updates are more likely to land in the root container; the sharder will deal with them as misplaced objects, and * shard containers created by new code on servers running old code will think they are root containers until the server is running new code, too; during this time they'll fail the sharder audit and report stats to their account, but both of these should get cleared up upon upgrade. Drive-by: fix a "conainer_name" typo that prevented us from testing that we can shard a container with unicode in its name. Also, add more UTF8 probe tests. [0] See https://bugs.python.org/issue22928 Change-Id: Ie08f36e31a448a547468dd85911c3a3bc30e89f1 Closes-Bug: 1856894
2019-12-18 15:14:00 -08:00
# We use native strings for metadata (see native_str_keys_and_values),
# so types are different.
if six.PY2:
key = u'test\u062a'.encode('utf-8')
sharding: Better-handle newlines in container names Previously, if you were on Python 2.7.10+ [0], such a newline would cause the sharder to fail, complaining about invalid header values when trying to create the shard containers. On older versions of Python, it would most likely cause a parsing error in the container-server that was trying to handle the PUT. Now, quote all places that we pass around container paths. This includes: * The X-Container-Sysmeta-Shard-(Quoted-)Root sent when creating the (empty) remote shards * The X-Container-Sysmeta-Shard-(Quoted-)Root included when initializing the local handoff for cleaving * The X-Backend-(Quoted-)Container-Path the proxy sends to the object-server for container updates * The Location header the container-server sends to the object-updater Note that a new header was required in requests so that servers would know whether the value should be unquoted or not. We can get away with reusing Location in responses by having clients opt-in to quoting with a new X-Backend-Accept-Quoted-Location header. During a rolling upgrade, * old object-servers servicing requests from new proxy-servers will not know about the container path override and so will try to update the root container, * in general, object updates are more likely to land in the root container; the sharder will deal with them as misplaced objects, and * shard containers created by new code on servers running old code will think they are root containers until the server is running new code, too; during this time they'll fail the sharder audit and report stats to their account, but both of these should get cleared up upon upgrade. Drive-by: fix a "conainer_name" typo that prevented us from testing that we can shard a container with unicode in its name. Also, add more UTF8 probe tests. [0] See https://bugs.python.org/issue22928 Change-Id: Ie08f36e31a448a547468dd85911c3a3bc30e89f1 Closes-Bug: 1856894
2019-12-18 15:14:00 -08:00
value = u'value\u062a'.encode('utf-8')
else:
key = u'test\u062a'
sharding: Better-handle newlines in container names Previously, if you were on Python 2.7.10+ [0], such a newline would cause the sharder to fail, complaining about invalid header values when trying to create the shard containers. On older versions of Python, it would most likely cause a parsing error in the container-server that was trying to handle the PUT. Now, quote all places that we pass around container paths. This includes: * The X-Container-Sysmeta-Shard-(Quoted-)Root sent when creating the (empty) remote shards * The X-Container-Sysmeta-Shard-(Quoted-)Root included when initializing the local handoff for cleaving * The X-Backend-(Quoted-)Container-Path the proxy sends to the object-server for container updates * The Location header the container-server sends to the object-updater Note that a new header was required in requests so that servers would know whether the value should be unquoted or not. We can get away with reusing Location in responses by having clients opt-in to quoting with a new X-Backend-Accept-Quoted-Location header. During a rolling upgrade, * old object-servers servicing requests from new proxy-servers will not know about the container path override and so will try to update the root container, * in general, object updates are more likely to land in the root container; the sharder will deal with them as misplaced objects, and * shard containers created by new code on servers running old code will think they are root containers until the server is running new code, too; during this time they'll fail the sharder audit and report stats to their account, but both of these should get cleared up upon upgrade. Drive-by: fix a "conainer_name" typo that prevented us from testing that we can shard a container with unicode in its name. Also, add more UTF8 probe tests. [0] See https://bugs.python.org/issue22928 Change-Id: Ie08f36e31a448a547468dd85911c3a3bc30e89f1 Closes-Bug: 1856894
2019-12-18 15:14:00 -08:00
value = u'value\u062a'
metadata = {
Add two vector timestamps The normalized form of the X-Timestamp header looks like a float with a fixed width to ensure stable string sorting - normalized timestamps look like "1402464677.04188" To support overwrites of existing data without modifying the original timestamp but still maintain consistency a second internal offset vector is append to the normalized timestamp form which compares and sorts greater than the fixed width float format but less than a newer timestamp. The internalized format of timestamps looks like "1402464677.04188_0000000000000000" - the portion after the underscore is the offset and is a formatted hexadecimal integer. The internalized form is not exposed to clients in responses from Swift. Normal client operations will not create a timestamp with an offset. The Timestamp class in common.utils supports internalized and normalized formatting of timestamps and also comparison of timestamp values. When the offset value of a Timestamp is 0 - it's considered insignificant and need not be represented in the string format; to support backwards compatibility during a Swift upgrade the internalized and normalized form of a Timestamp with an insignificant offset are identical. When a timestamp includes an offset it will always be represented in the internalized form, but is still excluded from the normalized form. Timestamps with an equivalent timestamp portion (the float part) will compare and order by their offset. Timestamps with a greater timestamp portion will always compare and order greater than a Timestamp with a lesser timestamp regardless of it's offset. String comparison and ordering is guaranteed for the internalized string format, and is backwards compatible for normalized timestamps which do not include an offset. The reconciler currently uses a offset bump to ensure that objects can move to the wrong storage policy and be moved back. This use-case is valid because the content represented by the user-facing timestamp is not modified in way. Future consumers of the offset vector of timestamps should be mindful of HTTP semantics of If-Modified and take care to avoid deviation in the response from the object server without an accompanying change to the user facing timestamp. DocImpact Implements: blueprint storage-policies Change-Id: Id85c960b126ec919a481dc62469bf172b7fb8549
2014-06-10 22:17:47 -07:00
key: [value, Timestamp(1).internal]
}
broker.update_metadata(metadata)
self.assertEqual(broker.metadata, metadata)
self.assertEqual(broker.get_raw_metadata(),
json.dumps(metadata))
def test_put_timestamp(self):
broker = self.broker_class(self.db_path, account='a', container='c')
orig_put_timestamp = next(self.ts).internal
broker.initialize(orig_put_timestamp,
storage_policy_index=int(self.policy))
self.assertEqual(broker.get_info()['put_timestamp'],
orig_put_timestamp)
# put_timestamp equal - no change
broker.update_put_timestamp(orig_put_timestamp)
self.assertEqual(broker.get_info()['put_timestamp'],
orig_put_timestamp)
# put_timestamp newer - gets newer
newer_put_timestamp = next(self.ts).internal
broker.update_put_timestamp(newer_put_timestamp)
self.assertEqual(broker.get_info()['put_timestamp'],
newer_put_timestamp)
# put_timestamp older - no change
broker.update_put_timestamp(orig_put_timestamp)
self.assertEqual(broker.get_info()['put_timestamp'],
newer_put_timestamp)
def test_status_changed_at(self):
broker = self.broker_class(self.db_path, account='test', container='c')
put_timestamp = next(self.ts).internal
created_at = time.time()
with patch('swift.common.db.time.time', new=lambda: created_at):
broker.initialize(put_timestamp,
storage_policy_index=int(self.policy))
self.assertEqual(broker.get_info()['status_changed_at'],
put_timestamp)
self.assertEqual(broker.get_info()['created_at'],
Timestamp(created_at).internal)
status_changed_at = next(self.ts).internal
broker.update_status_changed_at(status_changed_at)
self.assertEqual(broker.get_info()['status_changed_at'],
status_changed_at)
# save the old and get a new status_changed_at
old_status_changed_at, status_changed_at = \
status_changed_at, next(self.ts).internal
broker.update_status_changed_at(status_changed_at)
self.assertEqual(broker.get_info()['status_changed_at'],
status_changed_at)
# status changed at won't go backwards...
broker.update_status_changed_at(old_status_changed_at)
self.assertEqual(broker.get_info()['status_changed_at'],
status_changed_at)
def test_get_syncs(self):
broker = self.broker_class(self.db_path, account='a', container='c')
broker.initialize(Timestamp.now().internal,
storage_policy_index=int(self.policy))
self.assertEqual([], broker.get_syncs())
broker.merge_syncs([{'sync_point': 1, 'remote_id': 'remote1'}])
self.assertEqual([{'sync_point': 1, 'remote_id': 'remote1'}],
broker.get_syncs())
self.assertEqual([], broker.get_syncs(incoming=False))
broker.merge_syncs([{'sync_point': 2, 'remote_id': 'remote2'}],
incoming=False)
self.assertEqual([{'sync_point': 2, 'remote_id': 'remote2'}],
broker.get_syncs(incoming=False))
def test_commit_pending(self):
broker = self.broker_class(os.path.join(self.testdir, 'test.db'),
account='a', container='c')
broker.initialize(next(self.ts).internal,
storage_policy_index=int(self.policy))
self.put_item(broker, next(self.ts).internal)
qry = 'select * from %s_stat' % broker.db_type
with broker.get() as conn:
rows = [dict(x) for x in conn.execute(qry)]
info = rows[0]
count_key = '%s_count' % broker.db_contains_type
self.assertEqual(0, info[count_key])
# commit pending file into db
broker._commit_puts()
self.assertEqual(1, broker.get_info()[count_key])
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
def test_maybe_get(self):
broker = self.broker_class(os.path.join(self.testdir, 'test.db'),
account='a', container='c')
broker.initialize(next(self.ts).internal,
storage_policy_index=int(self.policy))
qry = 'select account from %s_stat' % broker.db_type
with broker.maybe_get(None) as conn:
rows = [dict(x) for x in conn.execute(qry)]
self.assertEqual([{'account': 'a'}], rows)
self.assertEqual(conn, broker.conn)
with broker.get() as other_conn:
self.assertEqual(broker.conn, None)
with broker.maybe_get(other_conn) as identity_conn:
self.assertIs(other_conn, identity_conn)
self.assertEqual(broker.conn, None)
self.assertEqual(broker.conn, None)
self.assertEqual(broker.conn, conn)
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
class TestDatabaseBroker(TestDbBase):
def test_DB_PREALLOCATION_setting(self):
u = uuid4().hex
b = DatabaseBroker(u)
swift.common.db.DB_PREALLOCATION = False
b._preallocate()
!! Changed db_preallocation to False Long explanation, but hopefully answers any questions. We don't like changing the default behavior of Swift unless there's a really good reason and, up until now, I've tried doing this with this new db_preallocation setting. For clusters with dedicated account/container servers that usually have fewer disks overall but SSD for speed, having db_preallocation on will gobble up disk space quite quickly and the fragmentation it's designed to fight isn't that big a speed impact to SSDs anyway. For clusters with account/container servers spread across all servers along with object servers usually having standard disks for cost, having db_preallocation off will cause very fragmented database files impacting speed, sometimes dramatically. Weighing these two negatives, it seems the second is the lesser evil. The first can cause disks to fill up and disable the cluster. The second will cause performance degradation, but the cluster will still function. Furthermore, if just one piece of code that touches all databases runs with db_preallocation on, it's effectively on for the whole cluster. We discovered this most recently when we finally configured everything within the Swift codebase to have db_preallocation off, only to find out Slogging didn't know about the new setting and so ran with it on and starting filling up SSDs. So that's why I'm proposing this change to the default behavior. We will definitely need to post a prominent notice of this change with the next release. Change-Id: I48a43439264cff5d03c14ec8787f718ee44e78ea
2012-05-22 00:30:47 +00:00
swift.common.db.DB_PREALLOCATION = True
self.assertRaises(OSError, b._preallocate)
2010-07-12 17:03:45 -05:00
def test_memory_db_init(self):
broker = DatabaseBroker(self.db_path)
self.assertEqual(broker.db_file, self.db_path)
2010-07-12 17:03:45 -05:00
self.assertRaises(AttributeError, broker.initialize,
normalize_timestamp('0'))
def test_disk_db_init(self):
db_file = os.path.join(self.testdir, '1.db')
broker = DatabaseBroker(db_file)
self.assertEqual(broker.db_file, db_file)
self.assertIsNone(broker.conn)
2010-07-12 17:03:45 -05:00
def test_disk_preallocate(self):
test_size = [-1]
def fallocate_stub(fd, size):
test_size[0] = size
with patch('swift.common.db.fallocate', fallocate_stub):
db_file = os.path.join(self.testdir, 'pre.db')
# Write 1 byte and hope that the fs will allocate less than 1 MB.
f = open(db_file, "w")
f.write('@')
f.close()
b = DatabaseBroker(db_file)
b._preallocate()
# We only wrote 1 byte, so we should end with the 1st step or 1 MB.
self.assertEqual(test_size[0], 1024 * 1024)
2010-07-12 17:03:45 -05:00
def test_initialize(self):
self.assertRaises(AttributeError,
DatabaseBroker(self.db_path).initialize,
2010-07-12 17:03:45 -05:00
normalize_timestamp('1'))
stub_dict = {}
2010-07-12 17:03:45 -05:00
def stub(*args, **kwargs):
stub_dict.clear()
2010-07-12 17:03:45 -05:00
stub_dict['args'] = args
stub_dict.update(kwargs)
broker = DatabaseBroker(self.db_path)
2010-07-12 17:03:45 -05:00
broker._initialize = stub
broker.initialize(normalize_timestamp('1'))
self.assertTrue(hasattr(stub_dict['args'][0], 'execute'))
self.assertEqual(stub_dict['args'][1], '0000000001.00000')
2010-07-12 17:03:45 -05:00
with broker.get() as conn:
conn.execute('SELECT * FROM outgoing_sync')
conn.execute('SELECT * FROM incoming_sync')
broker = DatabaseBroker(os.path.join(self.testdir, '1.db'))
broker._initialize = stub
broker.initialize(normalize_timestamp('1'))
self.assertTrue(hasattr(stub_dict['args'][0], 'execute'))
self.assertEqual(stub_dict['args'][1], '0000000001.00000')
2010-07-12 17:03:45 -05:00
with broker.get() as conn:
conn.execute('SELECT * FROM outgoing_sync')
conn.execute('SELECT * FROM incoming_sync')
broker = DatabaseBroker(os.path.join(self.testdir, '1.db'))
broker._initialize = stub
self.assertRaises(DatabaseAlreadyExists,
broker.initialize, normalize_timestamp('1'))
2010-07-12 17:03:45 -05:00
def test_delete_db(self):
meta = {'foo': ['bar', normalize_timestamp('0')]}
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
def init_stub(conn, put_timestamp, **kwargs):
2011-08-29 13:04:20 -05:00
conn.execute('CREATE TABLE test (one TEXT)')
conn.execute('''CREATE TABLE test_stat (
id TEXT, put_timestamp TEXT, delete_timestamp TEXT,
status TEXT, status_changed_at TEXT, metadata TEXT)''')
conn.execute(
'''INSERT INTO test_stat (
id, put_timestamp, delete_timestamp, status,
status_changed_at, metadata) VALUES (?, ?, ?, ?, ?, ?)''',
(str(uuid4), put_timestamp, '0', '', '0', json.dumps(meta)))
2011-08-29 13:04:20 -05:00
conn.execute('INSERT INTO test (one) VALUES ("1")')
conn.commit()
def do_test(expected_metadata, delete_meta_whitelist=None):
if not delete_meta_whitelist:
delete_meta_whitelist = []
broker = DatabaseBroker(self.get_db_path())
broker.delete_meta_whitelist = delete_meta_whitelist
broker.db_type = 'test'
broker._initialize = init_stub
# Initializes a good broker for us
broker.initialize(normalize_timestamp('1'))
info = broker.get_info()
self.assertEqual('0', info['delete_timestamp'])
self.assertEqual('', info['status'])
self.assertIsNotNone(broker.conn)
broker.delete_db(normalize_timestamp('2'))
info = broker.get_info()
self.assertEqual(normalize_timestamp('2'),
info['delete_timestamp'])
self.assertEqual('DELETED', info['status'])
# check meta
m2 = broker.metadata
self.assertEqual(m2, expected_metadata)
broker = DatabaseBroker(os.path.join(self.testdir,
'%s.db' % uuid4()))
broker.delete_meta_whitelist = delete_meta_whitelist
broker.db_type = 'test'
broker._initialize = init_stub
broker.initialize(normalize_timestamp('1'))
info = broker.get_info()
self.assertEqual('0', info['delete_timestamp'])
self.assertEqual('', info['status'])
broker.delete_db(normalize_timestamp('2'))
info = broker.get_info()
self.assertEqual(normalize_timestamp('2'),
info['delete_timestamp'])
self.assertEqual('DELETED', info['status'])
# check meta
m2 = broker.metadata
self.assertEqual(m2, expected_metadata)
# ensure that metadata was cleared by default
do_test({'foo': ['', normalize_timestamp('2')]})
# If the meta is in the brokers delete_meta_whitelist it wont get
# cleared up
do_test(meta, ['foo'])
# delete_meta_whitelist things need to be in lower case, as the keys
# are lower()'ed before checked
meta["X-Container-Meta-Test"] = ['value', normalize_timestamp('0')]
meta["X-Something-else"] = ['other', normalize_timestamp('0')]
do_test({'foo': ['', normalize_timestamp('2')],
'X-Container-Meta-Test': ['value', normalize_timestamp('0')],
'X-Something-else': ['other', normalize_timestamp('0')]},
['x-container-meta-test', 'x-something-else'])
2010-07-12 17:03:45 -05:00
def test_get(self):
broker = DatabaseBroker(self.db_path)
with self.assertRaises(DatabaseConnectionError) as raised, \
broker.get() as conn:
conn.execute('SELECT 1')
self.assertEqual(
str(raised.exception),
"DB connection error (%s, 0):\nDB doesn't exist" % self.db_path)
2010-07-12 17:03:45 -05:00
broker = DatabaseBroker(os.path.join(self.testdir, '1.db'))
with self.assertRaises(DatabaseConnectionError) as raised, \
broker.get() as conn:
conn.execute('SELECT 1')
self.assertEqual(
str(raised.exception),
"DB connection error (%s, 0):\nDB doesn't exist" % broker.db_file)
2010-07-12 17:03:45 -05:00
def stub(*args, **kwargs):
pass
broker._initialize = stub
broker.initialize(normalize_timestamp('1'))
with broker.get() as conn:
conn.execute('CREATE TABLE test (one TEXT)')
try:
with broker.get() as conn:
conn.execute('INSERT INTO test (one) VALUES ("1")')
raise Exception('test')
conn.commit()
2011-01-26 14:31:33 -08:00
except Exception:
2010-07-12 17:03:45 -05:00
pass
broker = DatabaseBroker(os.path.join(self.testdir, '1.db'))
with broker.get() as conn:
self.assertEqual(
2010-07-12 17:03:45 -05:00
[r[0] for r in conn.execute('SELECT * FROM test')], [])
with broker.get() as conn:
conn.execute('INSERT INTO test (one) VALUES ("1")')
conn.commit()
broker = DatabaseBroker(os.path.join(self.testdir, '1.db'))
with broker.get() as conn:
self.assertEqual(
2010-07-12 17:03:45 -05:00
[r[0] for r in conn.execute('SELECT * FROM test')], ['1'])
dbpath = os.path.join(self.testdir, 'dev', 'dbs', 'par', 'pre', 'db')
mkdirs(dbpath)
qpath = os.path.join(self.testdir, 'dev', 'quarantined', 'tests', 'db')
with patch('swift.common.db.renamer', lambda a, b,
fsync: b):
2011-08-01 20:46:30 +00:00
# Test malformed database
2011-08-02 18:21:25 +00:00
copy(os.path.join(os.path.dirname(__file__),
'malformed_example.db'),
os.path.join(dbpath, '1.db'))
broker = DatabaseBroker(os.path.join(dbpath, '1.db'))
2011-08-01 20:46:30 +00:00
broker.db_type = 'test'
with self.assertRaises(sqlite3.DatabaseError) as raised, \
broker.get() as conn:
conn.execute('SELECT * FROM test')
self.assertEqual(
str(raised.exception),
'Quarantined %s to %s due to malformed database' %
(dbpath, qpath))
# Test malformed schema database
copy(os.path.join(os.path.dirname(__file__),
'malformed_schema_example.db'),
os.path.join(dbpath, '1.db'))
broker = DatabaseBroker(os.path.join(dbpath, '1.db'))
broker.db_type = 'test'
with self.assertRaises(sqlite3.DatabaseError) as raised, \
broker.get() as conn:
conn.execute('SELECT * FROM test')
self.assertEqual(
str(raised.exception),
2011-08-01 21:14:41 +00:00
'Quarantined %s to %s due to malformed database' %
(dbpath, qpath))
2011-08-01 20:46:30 +00:00
# Test corrupted database
2011-08-02 18:21:25 +00:00
copy(os.path.join(os.path.dirname(__file__),
'corrupted_example.db'),
os.path.join(dbpath, '1.db'))
broker = DatabaseBroker(os.path.join(dbpath, '1.db'))
2011-08-01 20:46:30 +00:00
broker.db_type = 'test'
with self.assertRaises(sqlite3.DatabaseError) as raised, \
broker.get() as conn:
conn.execute('SELECT * FROM test')
self.assertEqual(
str(raised.exception),
2011-08-01 21:14:41 +00:00
'Quarantined %s to %s due to corrupted database' %
(dbpath, qpath))
2010-07-12 17:03:45 -05:00
def test_get_raw_metadata_missing_container_info(self):
# Test missing container_info/container_stat row
dbpath = os.path.join(self.testdir, 'dev', 'dbs', 'par', 'pre', 'db')
mkdirs(dbpath)
qpath = os.path.join(self.testdir, 'dev', 'quarantined', 'containers',
'db')
copy(os.path.join(os.path.dirname(__file__),
'missing_container_info.db'),
os.path.join(dbpath, '1.db'))
broker = DatabaseBroker(os.path.join(dbpath, '1.db'))
broker.db_type = 'container'
with self.assertRaises(sqlite3.DatabaseError) as raised:
broker.get_raw_metadata()
self.assertEqual(
str(raised.exception),
'Quarantined %s to %s due to missing row in container_stat table' %
(dbpath, qpath))
2010-07-12 17:03:45 -05:00
def test_lock(self):
broker = DatabaseBroker(os.path.join(self.testdir, '1.db'), timeout=.1)
with self.assertRaises(DatabaseConnectionError) as raised, \
broker.lock():
pass
self.assertEqual(
str(raised.exception),
"DB connection error (%s, 0):\nDB doesn't exist" % broker.db_file)
2010-07-12 17:03:45 -05:00
def stub(*args, **kwargs):
pass
broker._initialize = stub
broker.initialize(normalize_timestamp('1'))
with broker.lock():
pass
with broker.lock():
pass
with self.assertRaises(RuntimeError) as raised, broker.lock():
raise RuntimeError('boom!')
self.assertEqual(raised.exception.args[0], 'boom!')
broker2 = DatabaseBroker(os.path.join(self.testdir, '1.db'),
timeout=.1)
2010-07-12 17:03:45 -05:00
broker2._initialize = stub
with broker.lock():
# broker2 raises the timeout
with self.assertRaises(LockTimeout) as raised:
with broker2.lock():
pass
self.assertEqual(str(raised.exception),
'0.1 seconds: %s' % broker.db_file)
# and the timeout bubbles up out of broker.lock()
with self.assertRaises(LockTimeout) as raised:
with broker.lock():
with broker2.lock():
pass
self.assertEqual(str(raised.exception),
'0.1 seconds: %s' % broker.db_file)
2010-07-12 17:03:45 -05:00
try:
with broker.lock():
raise Exception('test')
2011-01-26 14:31:33 -08:00
except Exception:
2010-07-12 17:03:45 -05:00
pass
with broker.lock():
pass
def test_newid(self):
broker = DatabaseBroker(self.db_path)
2010-07-12 17:03:45 -05:00
broker.db_type = 'test'
broker.db_contains_type = 'test'
uuid1 = str(uuid4())
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
def _initialize(conn, timestamp, **kwargs):
2010-07-12 17:03:45 -05:00
conn.execute('CREATE TABLE test (one TEXT)')
conn.execute('CREATE TABLE test_stat (id TEXT)')
conn.execute('INSERT INTO test_stat (id) VALUES (?)', (uuid1,))
conn.commit()
broker._initialize = _initialize
broker.initialize(normalize_timestamp('1'))
uuid2 = str(uuid4())
broker.newid(uuid2)
with broker.get() as conn:
uuids = [r[0] for r in conn.execute('SELECT * FROM test_stat')]
self.assertEqual(len(uuids), 1)
self.assertNotEqual(uuids[0], uuid1)
2010-07-12 17:03:45 -05:00
uuid1 = uuids[0]
points = [(r[0], r[1]) for r in conn.execute(
'SELECT sync_point, '
2010-07-12 17:03:45 -05:00
'remote_id FROM incoming_sync WHERE remote_id = ?', (uuid2,))]
self.assertEqual(len(points), 1)
self.assertEqual(points[0][0], -1)
self.assertEqual(points[0][1], uuid2)
2010-07-12 17:03:45 -05:00
conn.execute('INSERT INTO test (one) VALUES ("1")')
conn.commit()
uuid3 = str(uuid4())
broker.newid(uuid3)
with broker.get() as conn:
uuids = [r[0] for r in conn.execute('SELECT * FROM test_stat')]
self.assertEqual(len(uuids), 1)
self.assertNotEqual(uuids[0], uuid1)
2010-07-12 17:03:45 -05:00
uuid1 = uuids[0]
points = [(r[0], r[1]) for r in conn.execute(
'SELECT sync_point, '
2010-07-12 17:03:45 -05:00
'remote_id FROM incoming_sync WHERE remote_id = ?', (uuid3,))]
self.assertEqual(len(points), 1)
self.assertEqual(points[0][1], uuid3)
2010-07-12 17:03:45 -05:00
broker.newid(uuid2)
with broker.get() as conn:
uuids = [r[0] for r in conn.execute('SELECT * FROM test_stat')]
self.assertEqual(len(uuids), 1)
self.assertNotEqual(uuids[0], uuid1)
points = [(r[0], r[1]) for r in conn.execute(
'SELECT sync_point, '
2010-07-12 17:03:45 -05:00
'remote_id FROM incoming_sync WHERE remote_id = ?', (uuid2,))]
self.assertEqual(len(points), 1)
self.assertEqual(points[0][1], uuid2)
2010-07-12 17:03:45 -05:00
def test_get_items_since(self):
broker = DatabaseBroker(self.db_path)
2010-07-12 17:03:45 -05:00
broker.db_type = 'test'
broker.db_contains_type = 'test'
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
def _initialize(conn, timestamp, **kwargs):
2010-07-12 17:03:45 -05:00
conn.execute('CREATE TABLE test (one TEXT)')
conn.execute('INSERT INTO test (one) VALUES ("1")')
conn.execute('INSERT INTO test (one) VALUES ("2")')
conn.execute('INSERT INTO test (one) VALUES ("3")')
conn.commit()
broker._initialize = _initialize
broker.initialize(normalize_timestamp('1'))
self.assertEqual(broker.get_items_since(-1, 10),
[{'one': '1'}, {'one': '2'}, {'one': '3'}])
self.assertEqual(broker.get_items_since(-1, 2),
[{'one': '1'}, {'one': '2'}])
self.assertEqual(broker.get_items_since(1, 2),
[{'one': '2'}, {'one': '3'}])
self.assertEqual(broker.get_items_since(3, 2), [])
self.assertEqual(broker.get_items_since(999, 2), [])
2010-07-12 17:03:45 -05:00
cli: add --sync to db info to show syncs When looking at containers and accounts it's sometimes nice to know who they've been replicating with. This patch adds a `--sync|-s` option to swift-{container|account}-info which will also dump the incoming and outgoing sync tables: $ swift-container-info /srv/node3/sdb3/containers/294/624/49b9ff074c502ec5e429e7af99a30624/49b9ff074c502ec5e429e7af99a30624.db -s Path: /AUTH_test/new Account: AUTH_test Container: new Deleted: False Container Hash: 49b9ff074c502ec5e429e7af99a30624 Metadata: Created at: 2022-02-16T05:34:05.988480 (1644989645.98848) Put Timestamp: 2022-02-16T05:34:05.981320 (1644989645.98132) Delete Timestamp: 1970-01-01T00:00:00.000000 (0) Status Timestamp: 2022-02-16T05:34:05.981320 (1644989645.98132) Object Count: 1 Bytes Used: 7 Storage Policy: default (0) Reported Put Timestamp: 1970-01-01T00:00:00.000000 (0) Reported Delete Timestamp: 1970-01-01T00:00:00.000000 (0) Reported Object Count: 0 Reported Bytes Used: 0 Chexor: 962368324c2ca023c56669d03ed92807 UUID: f33184e7-56d5-4c74-9d2e-5417c187d722-sdb3 X-Container-Sync-Point2: -1 X-Container-Sync-Point1: -1 No system metadata found in db file No user metadata found in db file Sharding Metadata: Type: root State: unsharded Incoming Syncs: Sync Point Remote ID Updated At 1 ce7268a1-f5d0-4b83-b993-af17b602a0ff-sdb1 2022-02-16T05:38:22.000000 (1644989902) 1 2af5abc0-7f70-4e2f-8f94-737aeaada7f4-sdb4 2022-02-16T05:38:22.000000 (1644989902) Outgoing Syncs: Sync Point Remote ID Updated At Partition 294 Hash 49b9ff074c502ec5e429e7af99a30624 As a follow up to the device in DB ID patch we can see that the replicas at sdb1 and sdb4 have replicated with this node. Change-Id: I23d786e82c6710bea7660a9acf8bbbd113b5b727
2022-02-16 16:38:52 +11:00
def test_get_syncs(self):
broker = DatabaseBroker(self.db_path)
broker.db_type = 'test'
broker.db_contains_type = 'test'
uuid1 = str(uuid4())
def _initialize(conn, timestamp, **kwargs):
conn.execute('CREATE TABLE test (one TEXT)')
conn.execute('CREATE TABLE test_stat (id TEXT)')
conn.execute('INSERT INTO test_stat (id) VALUES (?)', (uuid1,))
conn.execute('INSERT INTO test (one) VALUES ("1")')
conn.commit()
pass
broker._initialize = _initialize
broker.initialize(normalize_timestamp('1'))
for incoming in (True, False):
# Can't mock out timestamp now, because the update_at in the sync
# tables are cuase by a trigger inside sqlite which uses it's own
# now method. So instead track the time before and after to make
# sure we're getting the right timestamps.
ts0 = Timestamp.now()
broker.merge_syncs([
{'sync_point': 0, 'remote_id': 'remote_0'},
{'sync_point': 1, 'remote_id': 'remote_1'}], incoming)
time.sleep(0.005)
broker.merge_syncs([
{'sync_point': 2, 'remote_id': 'remote_2'}], incoming)
ts1 = Timestamp.now()
expected_syncs = [{'sync_point': 0, 'remote_id': 'remote_0'},
{'sync_point': 1, 'remote_id': 'remote_1'},
{'sync_point': 2, 'remote_id': 'remote_2'}]
self.assertEqual(expected_syncs, broker.get_syncs(incoming))
# if we want the updated_at timestamps too then:
expected_syncs[0]['updated_at'] = mock.ANY
expected_syncs[1]['updated_at'] = mock.ANY
expected_syncs[2]['updated_at'] = mock.ANY
actual_syncs = broker.get_syncs(incoming, include_timestamp=True)
self.assertEqual(expected_syncs, actual_syncs)
# Note that most of the time, we expect these all to be ==
# but we've been known to see sizeable delays in the gate at times
self.assertTrue(all([
str(int(ts0)) <= s['updated_at'] <= str(int(ts1))
for s in actual_syncs]))
2010-07-12 17:03:45 -05:00
def test_get_sync(self):
broker = DatabaseBroker(self.db_path)
2010-07-12 17:03:45 -05:00
broker.db_type = 'test'
broker.db_contains_type = 'test'
uuid1 = str(uuid4())
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
def _initialize(conn, timestamp, **kwargs):
2010-07-12 17:03:45 -05:00
conn.execute('CREATE TABLE test (one TEXT)')
conn.execute('CREATE TABLE test_stat (id TEXT)')
conn.execute('INSERT INTO test_stat (id) VALUES (?)', (uuid1,))
conn.execute('INSERT INTO test (one) VALUES ("1")')
conn.commit()
pass
broker._initialize = _initialize
broker.initialize(normalize_timestamp('1'))
uuid2 = str(uuid4())
self.assertEqual(broker.get_sync(uuid2), -1)
2010-07-12 17:03:45 -05:00
broker.newid(uuid2)
self.assertEqual(broker.get_sync(uuid2), 1)
2010-07-12 17:03:45 -05:00
uuid3 = str(uuid4())
self.assertEqual(broker.get_sync(uuid3), -1)
2010-07-12 17:03:45 -05:00
with broker.get() as conn:
conn.execute('INSERT INTO test (one) VALUES ("2")')
conn.commit()
broker.newid(uuid3)
self.assertEqual(broker.get_sync(uuid2), 1)
self.assertEqual(broker.get_sync(uuid3), 2)
self.assertEqual(broker.get_sync(uuid2, incoming=False), -1)
self.assertEqual(broker.get_sync(uuid3, incoming=False), -1)
2010-07-12 17:03:45 -05:00
broker.merge_syncs([{'sync_point': 1, 'remote_id': uuid2}],
incoming=False)
self.assertEqual(broker.get_sync(uuid2), 1)
self.assertEqual(broker.get_sync(uuid3), 2)
self.assertEqual(broker.get_sync(uuid2, incoming=False), 1)
self.assertEqual(broker.get_sync(uuid3, incoming=False), -1)
2010-07-12 17:03:45 -05:00
broker.merge_syncs([{'sync_point': 2, 'remote_id': uuid3}],
incoming=False)
self.assertEqual(broker.get_sync(uuid2, incoming=False), 1)
self.assertEqual(broker.get_sync(uuid3, incoming=False), 2)
2010-07-12 17:03:45 -05:00
def test_merge_syncs(self):
broker = DatabaseBroker(self.db_path)
2010-07-12 17:03:45 -05:00
def stub(*args, **kwargs):
pass
broker._initialize = stub
broker.initialize(normalize_timestamp('1'))
uuid2 = str(uuid4())
broker.merge_syncs([{'sync_point': 1, 'remote_id': uuid2}])
self.assertEqual(broker.get_sync(uuid2), 1)
2010-07-12 17:03:45 -05:00
uuid3 = str(uuid4())
broker.merge_syncs([{'sync_point': 2, 'remote_id': uuid3}])
self.assertEqual(broker.get_sync(uuid2), 1)
self.assertEqual(broker.get_sync(uuid3), 2)
self.assertEqual(broker.get_sync(uuid2, incoming=False), -1)
self.assertEqual(broker.get_sync(uuid3, incoming=False), -1)
2010-07-12 17:03:45 -05:00
broker.merge_syncs([{'sync_point': 3, 'remote_id': uuid2},
{'sync_point': 4, 'remote_id': uuid3}],
incoming=False)
self.assertEqual(broker.get_sync(uuid2, incoming=False), 3)
self.assertEqual(broker.get_sync(uuid3, incoming=False), 4)
self.assertEqual(broker.get_sync(uuid2), 1)
self.assertEqual(broker.get_sync(uuid3), 2)
2010-07-12 17:03:45 -05:00
broker.merge_syncs([{'sync_point': 5, 'remote_id': uuid2}])
self.assertEqual(broker.get_sync(uuid2), 5)
# max sync point sticks
broker.merge_syncs([{'sync_point': 5, 'remote_id': uuid2}])
self.assertEqual(broker.get_sync(uuid2), 5)
self.assertEqual(broker.get_sync(uuid3), 2)
broker.merge_syncs([{'sync_point': 4, 'remote_id': uuid2}])
self.assertEqual(broker.get_sync(uuid2), 5)
self.assertEqual(broker.get_sync(uuid3), 2)
broker.merge_syncs([{'sync_point': -1, 'remote_id': uuid2},
{'sync_point': 3, 'remote_id': uuid3}])
self.assertEqual(broker.get_sync(uuid2), 5)
self.assertEqual(broker.get_sync(uuid3), 3)
self.assertEqual(broker.get_sync(uuid2, incoming=False), 3)
self.assertEqual(broker.get_sync(uuid3, incoming=False), 4)
2010-07-12 17:03:45 -05:00
def test_get_replication_info(self):
self.get_replication_info_tester(metadata=False)
def test_get_replication_info_with_metadata(self):
self.get_replication_info_tester(metadata=True)
def get_replication_info_tester(self, metadata=False):
broker = DatabaseBroker(self.db_path, account='a')
broker.db_type = 'test'
broker.db_contains_type = 'test'
broker.db_reclaim_timestamp = 'created_at'
broker_creation = normalize_timestamp(1)
broker_uuid = str(uuid4())
broker_metadata = metadata and json.dumps(
{'Test': ('Value', normalize_timestamp(1))}) or ''
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
def _initialize(conn, put_timestamp, **kwargs):
if put_timestamp is None:
put_timestamp = normalize_timestamp(0)
conn.executescript('''
CREATE TABLE test (
ROWID INTEGER PRIMARY KEY AUTOINCREMENT,
name TEXT UNIQUE,
created_at TEXT
);
CREATE TRIGGER test_insert AFTER INSERT ON test
BEGIN
UPDATE test_stat
SET test_count = test_count + 1,
hash = chexor(hash, new.name, new.created_at);
END;
CREATE TRIGGER test_update BEFORE UPDATE ON test
BEGIN
SELECT RAISE(FAIL,
'UPDATE not allowed; DELETE and INSERT');
END;
CREATE TRIGGER test_delete AFTER DELETE ON test
BEGIN
UPDATE test_stat
SET test_count = test_count - 1,
hash = chexor(hash, old.name, old.created_at);
END;
CREATE TABLE test_stat (
account TEXT,
created_at TEXT,
put_timestamp TEXT DEFAULT '0',
delete_timestamp TEXT DEFAULT '0',
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
status_changed_at TEXT DEFAULT '0',
test_count INTEGER,
hash TEXT default '00000000000000000000000000000000',
id TEXT
%s
);
INSERT INTO test_stat (test_count) VALUES (0);
''' % (metadata and ", metadata TEXT DEFAULT ''" or ""))
conn.execute('''
UPDATE test_stat
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
SET account = ?, created_at = ?, id = ?, put_timestamp = ?,
status_changed_at = ?
''', (broker.account, broker_creation, broker_uuid, put_timestamp,
put_timestamp))
if metadata:
conn.execute('UPDATE test_stat SET metadata = ?',
(broker_metadata,))
conn.commit()
broker._initialize = _initialize
put_timestamp = normalize_timestamp(2)
broker.initialize(put_timestamp)
info = broker.get_replication_info()
self.assertEqual(info, {
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
'account': broker.account, 'count': 0,
'hash': '00000000000000000000000000000000',
'created_at': broker_creation, 'put_timestamp': put_timestamp,
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
'delete_timestamp': '0', 'status_changed_at': put_timestamp,
'max_row': -1, 'id': broker_uuid, 'metadata': broker_metadata})
insert_timestamp = normalize_timestamp(3)
with broker.get() as conn:
conn.execute('''
INSERT INTO test (name, created_at) VALUES ('test', ?)
''', (insert_timestamp,))
conn.commit()
info = broker.get_replication_info()
self.assertEqual(info, {
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
'account': broker.account, 'count': 1,
'hash': 'bdc4c93f574b0d8c2911a27ce9dd38ba',
'created_at': broker_creation, 'put_timestamp': put_timestamp,
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
'delete_timestamp': '0', 'status_changed_at': put_timestamp,
'max_row': 1, 'id': broker_uuid, 'metadata': broker_metadata})
with broker.get() as conn:
conn.execute('DELETE FROM test')
conn.commit()
info = broker.get_replication_info()
self.assertEqual(info, {
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
'account': broker.account, 'count': 0,
'hash': '00000000000000000000000000000000',
'created_at': broker_creation, 'put_timestamp': put_timestamp,
Add Storage Policy support to Containers Containers now have a storage policy index associated with them, stored in the container_stat table. This index is only settable at container creation time (PUT request), and cannot be changed without deleting and recreating the container. This is because a container's policy index will apply to all its objects, so changing a container's policy index would require moving large amounts of object data around. If a user wants to change the policy for data in a container, they must create a new container with the desired policy and move the data over. Keep status_changed_at up-to-date with status changes. In particular during container recreation and replication. When a container-server receives a PUT for a deleted database an extra UPDATE is issued against the container_stat table to notate the x-timestamp of the request. During replication if merge_timestamps causes a container's status to change (from DELETED to ACTIVE or vice-versa) the status_changed_at field is set to the current time. Accurate reporting of status_changed_at is useful for container replication forensics and allows resolution of "set on create" attributes like the upcoming storage_policy_index. Expose Backend container info on deleted containers. Include basic container info in backend headers on 404 responses from the container server. Default empty values are used as placeholders if the database does not exist. Specifically the X-Backend-Status-Changed-At, X-Backend-DELETE-Timestamp and the X-Backend-Storage-Policy-Index value will be needed by the reconciler to deal with reconciling out of order object writes in the face of recently deleted containers. * Add "status_changed_at" key to the response from ContainerBroker.get_info. * Add "Status Timestamp" field to swift.cli.info.print_db_info_metadata. * Add "status_changed_at" key to the response from AccountBroker.get_info. DocImpact Implements: blueprint storage-policies Change-Id: Ie6d388f067f5b096b0f96faef151120ba23c8748
2014-05-27 16:57:25 -07:00
'delete_timestamp': '0', 'status_changed_at': put_timestamp,
'max_row': 1, 'id': broker_uuid, 'metadata': broker_metadata})
return broker
# only testing _reclaim_metadata here
@patch.object(TombstoneReclaimer, 'reclaim')
def test_metadata(self, mock_reclaim):
# Initializes a good broker for us
broker = self.get_replication_info_tester(metadata=True)
# Add our first item
first_timestamp = normalize_timestamp(1)
first_value = '1'
broker.update_metadata({'First': [first_value, first_timestamp]})
self.assertIn('First', broker.metadata)
self.assertEqual(broker.metadata['First'],
[first_value, first_timestamp])
# Add our second item
second_timestamp = normalize_timestamp(2)
second_value = '2'
broker.update_metadata({'Second': [second_value, second_timestamp]})
self.assertIn('First', broker.metadata)
self.assertEqual(broker.metadata['First'],
[first_value, first_timestamp])
self.assertIn('Second', broker.metadata)
self.assertEqual(broker.metadata['Second'],
[second_value, second_timestamp])
# Update our first item
first_timestamp = normalize_timestamp(3)
first_value = '1b'
broker.update_metadata({'First': [first_value, first_timestamp]})
self.assertIn('First', broker.metadata)
self.assertEqual(broker.metadata['First'],
[first_value, first_timestamp])
self.assertIn('Second', broker.metadata)
self.assertEqual(broker.metadata['Second'],
[second_value, second_timestamp])
# Delete our second item (by setting to empty string)
second_timestamp = normalize_timestamp(4)
second_value = ''
broker.update_metadata({'Second': [second_value, second_timestamp]})
self.assertIn('First', broker.metadata)
self.assertEqual(broker.metadata['First'],
[first_value, first_timestamp])
self.assertIn('Second', broker.metadata)
self.assertEqual(broker.metadata['Second'],
[second_value, second_timestamp])
# Reclaim at point before second item was deleted
broker.reclaim(normalize_timestamp(3), normalize_timestamp(3))
self.assertIn('First', broker.metadata)
self.assertEqual(broker.metadata['First'],
[first_value, first_timestamp])
self.assertIn('Second', broker.metadata)
self.assertEqual(broker.metadata['Second'],
[second_value, second_timestamp])
# Reclaim at point second item was deleted
broker.reclaim(normalize_timestamp(4), normalize_timestamp(4))
self.assertIn('First', broker.metadata)
self.assertEqual(broker.metadata['First'],
[first_value, first_timestamp])
self.assertIn('Second', broker.metadata)
self.assertEqual(broker.metadata['Second'],
[second_value, second_timestamp])
# Reclaim after point second item was deleted
broker.reclaim(normalize_timestamp(5), normalize_timestamp(5))
self.assertIn('First', broker.metadata)
self.assertEqual(broker.metadata['First'],
[first_value, first_timestamp])
self.assertNotIn('Second', broker.metadata)
# Delete first item (by setting to empty string)
first_timestamp = normalize_timestamp(6)
broker.update_metadata({'First': ['', first_timestamp]})
self.assertIn('First', broker.metadata)
# Check that sync_timestamp doesn't cause item to be reclaimed
broker.reclaim(normalize_timestamp(5), normalize_timestamp(99))
self.assertIn('First', broker.metadata)
def test_update_metadata_missing_container_info(self):
# Test missing container_info/container_stat row
dbpath = os.path.join(self.testdir, 'dev', 'dbs', 'par', 'pre', 'db')
mkdirs(dbpath)
qpath = os.path.join(self.testdir, 'dev', 'quarantined', 'containers',
'db')
copy(os.path.join(os.path.dirname(__file__),
'missing_container_info.db'),
os.path.join(dbpath, '1.db'))
broker = DatabaseBroker(os.path.join(dbpath, '1.db'))
broker.db_type = 'container'
with self.assertRaises(sqlite3.DatabaseError) as raised:
broker.update_metadata({'First': ['1', normalize_timestamp(1)]})
self.assertEqual(
str(raised.exception),
'Quarantined %s to %s due to missing row in container_stat table' %
(dbpath, qpath))
def test_reclaim_missing_container_info(self):
# Test missing container_info/container_stat row
dbpath = os.path.join(self.testdir, 'dev', 'dbs', 'par', 'pre', 'db')
mkdirs(dbpath)
qpath = os.path.join(self.testdir, 'dev', 'quarantined', 'containers',
'db')
copy(os.path.join(os.path.dirname(__file__),
'missing_container_info.db'),
os.path.join(dbpath, '1.db'))
broker = DatabaseBroker(os.path.join(dbpath, '1.db'))
broker.db_type = 'container'
with self.assertRaises(sqlite3.DatabaseError) as raised, \
broker.get() as conn:
broker._reclaim_metadata(conn, 0)
self.assertEqual(
str(raised.exception),
'Quarantined %s to %s due to missing row in container_stat table' %
(dbpath, qpath))
@patch.object(DatabaseBroker, 'validate_metadata')
def test_validate_metadata_is_called_from_update_metadata(self, mock):
broker = self.get_replication_info_tester(metadata=True)
first_timestamp = normalize_timestamp(1)
first_value = '1'
metadata = {'First': [first_value, first_timestamp]}
broker.update_metadata(metadata, validate_metadata=True)
self.assertTrue(mock.called)
@patch.object(DatabaseBroker, 'validate_metadata')
def test_validate_metadata_is_not_called_from_update_metadata(self, mock):
broker = self.get_replication_info_tester(metadata=True)
first_timestamp = normalize_timestamp(1)
first_value = '1'
metadata = {'First': [first_value, first_timestamp]}
broker.update_metadata(metadata)
self.assertFalse(mock.called)
def test_metadata_with_max_count(self):
metadata = {}
for c in range(MAX_META_COUNT):
key = 'X-Account-Meta-F{0}'.format(c)
metadata[key] = ('B', normalize_timestamp(1))
key = 'X-Account-Meta-Foo'
metadata[key] = ('', normalize_timestamp(1))
self.assertIsNone(DatabaseBroker.validate_metadata(metadata))
def test_metadata_raises_exception_on_non_utf8(self):
def try_validate(metadata):
with self.assertRaises(HTTPException) as raised:
DatabaseBroker.validate_metadata(metadata)
self.assertEqual(str(raised.exception), '400 Bad Request')
ts = normalize_timestamp(1)
try_validate({'X-Account-Meta-Foo': (b'\xff', ts)})
try_validate({b'X-Container-Meta-\xff': ('bar', ts)})
def test_metadata_raises_exception_over_max_count(self):
metadata = {}
for c in range(MAX_META_COUNT + 1):
key = 'X-Account-Meta-F{0}'.format(c)
metadata[key] = ('B', normalize_timestamp(1))
message = ''
try:
DatabaseBroker.validate_metadata(metadata)
except HTTPException as e:
message = str(e)
self.assertEqual(message, '400 Bad Request')
def test_metadata_with_max_overall_size(self):
metadata = {}
metadata_value = 'v' * MAX_META_VALUE_LENGTH
size = 0
x = 0
while size < (MAX_META_OVERALL_SIZE - 4
- MAX_META_VALUE_LENGTH):
size += 4 + MAX_META_VALUE_LENGTH
metadata['X-Account-Meta-%04d' % x] = (metadata_value,
normalize_timestamp(1))
x += 1
if MAX_META_OVERALL_SIZE - size > 1:
metadata['X-Account-Meta-k'] = (
'v' * (MAX_META_OVERALL_SIZE - size - 1),
normalize_timestamp(1))
self.assertIsNone(DatabaseBroker.validate_metadata(metadata))
def test_metadata_raises_exception_over_max_overall_size(self):
metadata = {}
metadata_value = 'k' * MAX_META_VALUE_LENGTH
size = 0
x = 0
while size < (MAX_META_OVERALL_SIZE - 4
- MAX_META_VALUE_LENGTH):
size += 4 + MAX_META_VALUE_LENGTH
metadata['X-Account-Meta-%04d' % x] = (metadata_value,
normalize_timestamp(1))
x += 1
if MAX_META_OVERALL_SIZE - size > 1:
metadata['X-Account-Meta-k'] = (
'v' * (MAX_META_OVERALL_SIZE - size - 1),
normalize_timestamp(1))
metadata['X-Account-Meta-k2'] = ('v', normalize_timestamp(1))
message = ''
try:
DatabaseBroker.validate_metadata(metadata)
except HTTPException as e:
message = str(e)
self.assertEqual(message, '400 Bad Request')
def test_possibly_quarantine_db_errors(self):
dbpath = os.path.join(self.testdir, 'dev', 'dbs', 'par', 'pre', 'db')
qpath = os.path.join(self.testdir, 'dev', 'quarantined', 'tests', 'db')
# Data is a list of Excpetions to be raised and expected values in the
# log
data = [
(sqlite3.DatabaseError('database disk image is malformed'),
'malformed'),
(sqlite3.DatabaseError('malformed database schema'), 'malformed'),
(sqlite3.DatabaseError('file is encrypted or is not a database'),
'corrupted'),
(sqlite3.OperationalError('disk I/O error'),
'disk error while accessing')]
for i, (ex, hint) in enumerate(data):
mkdirs(dbpath)
broker = DatabaseBroker(os.path.join(dbpath, '%d.db' % (i)))
broker.db_type = 'test'
try:
raise ex
except sqlite3.DatabaseError:
with self.assertRaises(sqlite3.DatabaseError) as raised:
broker.possibly_quarantine(*sys.exc_info())
self.assertEqual(
str(raised.exception),
'Quarantined %s to %s due to %s database' %
(dbpath, qpath, hint))
2010-07-12 17:03:45 -05:00
def test_skip_commits(self):
broker = DatabaseBroker(self.db_path)
self.assertTrue(broker._skip_commit_puts())
broker._initialize = MagicMock()
broker.initialize(Timestamp.now())
self.assertTrue(broker._skip_commit_puts())
# not initialized
db_file = os.path.join(self.testdir, '1.db')
broker = DatabaseBroker(db_file)
self.assertFalse(os.path.exists(broker.db_file)) # sanity check
self.assertTrue(broker._skip_commit_puts())
# no pending file
broker._initialize = MagicMock()
broker.initialize(Timestamp.now())
self.assertTrue(os.path.exists(broker.db_file)) # sanity check
self.assertFalse(os.path.exists(broker.pending_file)) # sanity check
self.assertTrue(broker._skip_commit_puts())
# pending file exists
with open(broker.pending_file, 'wb'):
pass
self.assertTrue(os.path.exists(broker.pending_file)) # sanity check
self.assertFalse(broker._skip_commit_puts())
# skip_commits is True
broker.skip_commits = True
self.assertTrue(broker._skip_commit_puts())
# re-init
broker = DatabaseBroker(db_file)
self.assertFalse(broker._skip_commit_puts())
# constructor can override
broker = DatabaseBroker(db_file, skip_commits=True)
self.assertTrue(broker._skip_commit_puts())
def test_commit_puts(self):
db_file = os.path.join(self.testdir, '1.db')
broker = DatabaseBroker(db_file)
broker._initialize = MagicMock()
broker.initialize(Timestamp.now())
with open(broker.pending_file, 'wb'):
pass
# merge given list
with patch.object(broker, 'merge_items') as mock_merge_items:
broker._commit_puts(['test'])
mock_merge_items.assert_called_once_with(['test'])
# load file and merge
with open(broker.pending_file, 'wb') as fd:
for v in (1, 2, 99):
fd.write(b':' + base64.b64encode(pickle.dumps(
v, protocol=PICKLE_PROTOCOL)))
with patch.object(broker, 'merge_items') as mock_merge_items:
broker._commit_puts_load = lambda l, e: l.append(e)
broker._commit_puts()
mock_merge_items.assert_called_once_with([1, 2, 99])
self.assertEqual(0, os.path.getsize(broker.pending_file))
# load file and merge with given list
with open(broker.pending_file, 'wb') as fd:
fd.write(b':' + base64.b64encode(pickle.dumps(
b'bad', protocol=PICKLE_PROTOCOL)))
with patch.object(broker, 'merge_items') as mock_merge_items:
broker._commit_puts_load = lambda l, e: l.append(e)
broker._commit_puts([b'not'])
mock_merge_items.assert_called_once_with([b'not', b'bad'])
self.assertEqual(0, os.path.getsize(broker.pending_file))
# load a pending entry that's caused trouble in py2/py3 upgrade tests
# can't quite figure out how it got generated, though, so hard-code it
with open(broker.pending_file, 'wb') as fd:
fd.write(b':gAIoVS3olIngpILrjIvrjIvpkIngpIHlmIjlmIbjnIbgp'
b'IPjnITimIPvhI/rjI3tiI5xAVUQMTU1OTI0MTg0Ni40NjY'
b'wMXECVQEwVQEwVQEwSwBVATB0Lg==')
with patch.object(broker, 'merge_items') as mock_merge_items:
broker._commit_puts_load = lambda l, e: l.append(e)
broker._commit_puts([])
expected_name = (u'\u8509\u0902\ub30b\ub30b\u9409\u0901\u5608\u5606'
u'\u3706\u0903\u3704\u2603\uf10f\ub30d\ud20e')
if six.PY2:
expected_name = expected_name.encode('utf8')
mock_merge_items.assert_called_once_with([
(expected_name, '1559241846.46601', '0', '0', '0', 0, '0')])
self.assertEqual(0, os.path.getsize(broker.pending_file))
# skip_commits True - no merge
db_file = os.path.join(self.testdir, '2.db')
broker = DatabaseBroker(db_file, skip_commits=True)
broker._initialize = MagicMock()
broker.initialize(Timestamp.now())
with open(broker.pending_file, 'wb') as fd:
fd.write(b':ignored')
with patch.object(broker, 'merge_items') as mock_merge_items:
with self.assertRaises(DatabaseConnectionError) as cm:
broker._commit_puts([b'hmmm'])
mock_merge_items.assert_not_called()
self.assertIn('commits not accepted', str(cm.exception))
with open(broker.pending_file, 'rb') as fd:
self.assertEqual(b':ignored', fd.read())
def test_put_record(self):
db_file = os.path.join(self.testdir, '1.db')
broker = DatabaseBroker(db_file)
broker._initialize = MagicMock()
broker.initialize(Timestamp.now())
# pending file created and record written
broker.make_tuple_for_pickle = lambda x: x.upper()
with patch.object(broker, '_commit_puts') as mock_commit_puts:
broker.put_record('pinky')
mock_commit_puts.assert_not_called()
with open(broker.pending_file, 'rb') as fd:
pending = fd.read()
items = pending.split(b':')
self.assertEqual(['PINKY'],
[pickle.loads(base64.b64decode(i))
for i in items[1:]])
# record appended
with patch.object(broker, '_commit_puts') as mock_commit_puts:
broker.put_record('perky')
mock_commit_puts.assert_not_called()
with open(broker.pending_file, 'rb') as fd:
pending = fd.read()
items = pending.split(b':')
self.assertEqual(['PINKY', 'PERKY'],
[pickle.loads(base64.b64decode(i))
for i in items[1:]])
# pending file above cap
cap = swift.common.db.PENDING_CAP
while os.path.getsize(broker.pending_file) < cap:
with open(broker.pending_file, 'ab') as fd:
fd.write(b'x' * 100000)
with patch.object(broker, '_commit_puts') as mock_commit_puts:
broker.put_record('direct')
mock_commit_puts.assert_called_once_with(['direct'])
# records shouldn't be put to brokers with skip_commits True because
# they cannot be accepted if the pending file is full
broker.skip_commits = True
with open(broker.pending_file, 'wb'):
# empty the pending file
pass
with patch.object(broker, '_commit_puts') as mock_commit_puts:
with self.assertRaises(DatabaseConnectionError) as cm:
broker.put_record('unwelcome')
self.assertIn('commits not accepted', str(cm.exception))
mock_commit_puts.assert_not_called()
with open(broker.pending_file, 'rb') as fd:
pending = fd.read()
self.assertFalse(pending)
class TestTombstoneReclaimer(TestDbBase):
def _make_object(self, broker, obj_name, ts, deleted):
if deleted:
broker.delete_test(obj_name, ts.internal)
else:
broker.put_test(obj_name, ts.internal)
def _count_reclaimable(self, conn, reclaim_age):
return conn.execute(
"SELECT count(*) FROM test "
"WHERE deleted = 1 AND created_at < ?", (reclaim_age,)
).fetchone()[0]
def _get_reclaimable(self, broker, reclaim_age):
with broker.get() as conn:
return self._count_reclaimable(conn, reclaim_age)
def _setup_tombstones(self, reverse_names=True):
broker = ExampleBroker(self.db_path,
account='test_account',
container='test_container')
broker.initialize(Timestamp('1').internal, 0)
now = time.time()
top_of_the_minute = now - (now % 60)
# namespace if reverse:
# a-* has 70 'active' tombstones followed by 70 reclaimable
# b-* has 70 'active' tombstones followed by 70 reclaimable
# else:
# a-* has 70 reclaimable followed by 70 'active' tombstones
# b-* has 70 reclaimable followed by 70 'active' tombstones
for i in range(0, 560, 4):
self._make_object(
broker, 'a_%3d' % (560 - i if reverse_names else i),
Timestamp(top_of_the_minute - (i * 60)), True)
self._make_object(
broker, 'a_%3d' % (559 - i if reverse_names else i + 1),
Timestamp(top_of_the_minute - ((i + 1) * 60)), False)
self._make_object(
broker, 'b_%3d' % (560 - i if reverse_names else i),
Timestamp(top_of_the_minute - ((i + 2) * 60)), True)
self._make_object(
broker, 'b_%3d' % (559 - i if reverse_names else i + 1),
Timestamp(top_of_the_minute - ((i + 3) * 60)), False)
broker._commit_puts()
# divide the set of timestamps exactly in half for reclaim
reclaim_age = top_of_the_minute + 1 - (560 / 2 * 60)
self.assertEqual(140, self._get_reclaimable(broker, reclaim_age))
tombstones = self._get_reclaimable(broker, top_of_the_minute + 1)
self.assertEqual(280, tombstones)
return broker, top_of_the_minute, reclaim_age
@contextlib.contextmanager
def _mock_broker_get(self, broker, reclaim_age):
# intercept broker.get() calls and capture the current reclaimable
# count before returning a conn
orig_get = broker.get
reclaimable = []
@contextlib.contextmanager
def mock_get():
with orig_get() as conn:
reclaimable.append(self._count_reclaimable(conn, reclaim_age))
yield conn
with patch.object(broker, 'get', mock_get):
yield reclaimable
def test_batched_reclaim_several_small_batches(self):
broker, totm, reclaim_age = self._setup_tombstones()
with self._mock_broker_get(broker, reclaim_age) as reclaimable:
with patch('swift.common.db.RECLAIM_PAGE_SIZE', 50):
reclaimer = TombstoneReclaimer(broker, reclaim_age)
reclaimer.reclaim()
expected_reclaimable = [140, # 0 rows fetched
90, # 50 rows fetched, 50 reclaimed
70, # 100 rows fetched, 20 reclaimed
60, # 150 rows fetched, 10 reclaimed
10, # 200 rows fetched, 50 reclaimed
0, # 250 rows fetched, 10 reclaimed
]
self.assertEqual(expected_reclaimable, reclaimable)
self.assertEqual(0, self._get_reclaimable(broker, reclaim_age))
def test_batched_reclaim_exactly_two_batches(self):
broker, totm, reclaim_age = self._setup_tombstones()
with self._mock_broker_get(broker, reclaim_age) as reclaimable:
with patch('swift.common.db.RECLAIM_PAGE_SIZE', 140):
reclaimer = TombstoneReclaimer(broker, reclaim_age)
reclaimer.reclaim()
expected_reclaimable = [140, # 0 rows fetched
70, # 140 rows fetched, 70 reclaimed
]
self.assertEqual(expected_reclaimable, reclaimable)
self.assertEqual(0, self._get_reclaimable(broker, reclaim_age))
def test_batched_reclaim_one_large_batch(self):
broker, totm, reclaim_age = self._setup_tombstones()
with self._mock_broker_get(broker, reclaim_age) as reclaimable:
with patch('swift.common.db.RECLAIM_PAGE_SIZE', 1000):
reclaimer = TombstoneReclaimer(broker, reclaim_age)
reclaimer.reclaim()
expected_reclaimable = [140] # 0 rows fetched
self.assertEqual(expected_reclaimable, reclaimable)
self.assertEqual(0, self._get_reclaimable(broker, reclaim_age))
def test_reclaim_get_tombstone_count(self):
broker, totm, reclaim_age = self._setup_tombstones(reverse_names=False)
with patch('swift.common.db.RECLAIM_PAGE_SIZE', 122):
reclaimer = TombstoneReclaimer(broker, reclaim_age)
reclaimer.reclaim()
self.assertEqual(0, self._get_reclaimable(broker, reclaim_age))
tombstones = self._get_reclaimable(broker, totm + 1)
self.assertEqual(140, tombstones)
# in this scenario the reclaim phase finds the remaining tombstone
# count (140)
self.assertEqual(140, reclaimer.remaining_tombstones)
self.assertEqual(140, reclaimer.get_tombstone_count())
def test_reclaim_get_tombstone_count_with_leftover(self):
broker, totm, reclaim_age = self._setup_tombstones()
with patch('swift.common.db.RECLAIM_PAGE_SIZE', 122):
reclaimer = TombstoneReclaimer(broker, reclaim_age)
reclaimer.reclaim()
self.assertEqual(0, self._get_reclaimable(broker, reclaim_age))
tombstones = self._get_reclaimable(broker, totm + 1)
self.assertEqual(140, tombstones)
# in this scenario the reclaim phase finds a subset (104) of all
# tombstones (140)
self.assertEqual(104, reclaimer.remaining_tombstones)
# get_tombstone_count finds the rest
actual = reclaimer.get_tombstone_count()
self.assertEqual(140, actual)
def test_get_tombstone_count_with_leftover(self):
# verify that a call to get_tombstone_count() will invoke a reclaim if
# reclaim not already invoked
broker, totm, reclaim_age = self._setup_tombstones()
with patch('swift.common.db.RECLAIM_PAGE_SIZE', 122):
reclaimer = TombstoneReclaimer(broker, reclaim_age)
actual = reclaimer.get_tombstone_count()
self.assertEqual(0, self._get_reclaimable(broker, reclaim_age))
self.assertEqual(140, actual)
2010-07-12 17:03:45 -05:00
if __name__ == '__main__':
unittest.main()