CHANGES¶

1.57.4¶

• Switch from oslosphinx to openstackdocstheme
• Turn on warning-is-error in doc build

1.57.3¶

• Make sure Lock.heartbeat() returns True/False
• etcd3: skip ProcessPool based test
• etcd3: replace custom lock code by more recent etcd3 lock code
• pgsql: fix self._conn.close() being called without connection
• test: leverage existing helper method in test_partitioner
• coordination: remove double serialization of capabilities
• tests: fix missing .get() on some group operations
• Mutualize executor code in a mixin class
• coordination: fix reversed fiels for __repr__ for events
• Fix docstring for group and member id

1.57.2¶

• {my,pg}sql: close connections when out of retry
• Disable test_get_lock_serial_locking_two_lock_process for etcd3
• Simplify env list and test running
• Factorize tox envlist for better readability

1.57.1¶

• consul: remove unused executor

1.57.0¶

• Separate etcd3gw driver that uses the etcd3 grpc gateway
• etcd3: use discard() rather than remove()
• etcd: fix blocking argument
• etcd: fix acquire(blocking=True) on request exception
• etcd3: fix test run
• coordination: factorize common async result futures code

1.56.1¶

• Fix psycopg2 connection argument

1.56.0¶

• doc: update heartbeat doc to use start_heart=True
• sql: close connection for lock if not used
• http->https for security
• etcd3: add etcd3 coordination driver

1.55.0¶

• Implement heartbeat for FileDriver
• simplify hashring node lookup

1.54.0¶

• redis: fix concurrent access on acquire()
• tests: tests fail if no URL is set + run partitioner tests on basic drivers
• tests: fix etcd and consul test run
• add weight tests for add_nodes
• get weight of existing members
• coordination: do not get member list if not needed
• Add shared filelock

1.53.0¶

• Enhance heartbeat sleep timer

1.52.0¶

• FileDriver:Support multiple processes
• Switch tests to use latest etcd - 3.1.3

1.51.0¶

• pass on partitions
• hashring: allow to use bytes as node name

1.50.0¶

• postgresql: only pass username and password if they are set
• Rewrite heartbeat runner with event
• Adds authentication support for zookeeperDriver

1.49.0¶

• support unicode node name
• Update reno for stable/ocata

1.48.0¶

• Fix test function name with two underscores to have only one

1.47.0¶

• Add partitioner implementation
• Stop making tooz.utils depending on tooz.coordination
• [doc] Note lack of constraints is a choice
• The ‘moves.moved_class’ function creates a new class
• Add weight support to the hashring
• coordination: allow to pass capabilities in join_group_create()
• coordination: fix moved_class usage for ToozError
• zookeeper: switch to standard group membership watching
• Add a hashring implementation
• Move ToozError to root module
• Fixup concurrent modification
• Replaces uuid.uuid4 with uuidutils.generate_uuid()

1.46.0¶

• Do not re-set the members cache for watchers by default
• coordination: add __repr__ for join/leave events
• coordination: renforce event based testing
• Factorize member_id in the base coordinator class
• Use the internal group of list rather than listing the groups
• Move the cached-based watcher implementation into its own class

1.45.0¶

• Factorize group quit on stop()
• coordination: make get_members() return a set
• Replace ‘assertTrue(a (not)in b)’ with ‘assert(Not)In(a, b)’
• Changed author and author-email
• redis: make sure we don’t release and heartbeat a lock at the same time
• coordinator: add join_group_create
• Replace retrying with tenacity
• Add CONTRIBUTING.rst
• Replace ‘assertTrue(a in b)’ with ‘assertIn(a, b)’ and ‘assertFalse(a in b)’ with ‘assertNotIn(a, b)’
• Using assertIsNone() instead of assertEqual(None, …)
• tox: use pretty tox output
• tox: install docs dependency in docs target and reno
• Bump hacking to 0.12
• Add reno for release notes management

1.44.0¶

• Changed the home-page link
• file: update .metadata atomically
• file: return converted voluptuous data
• file: move _load_and_validate to a method
• file: move _read_{group,member}_id to staticmethod-s
• etcd: run tests in clustering mode too
• Use method ensure_tree from oslo.utils
• Switch from Python 3.4 to Python 3.5
• Install only needed packages
• Fix a typo in file.py
• Update etcd version in tests

1.43.0¶

• Makedirs only throws oserror, so only catch that

1.42.0¶

• etcd: don’t run heartbeat() concurrently
• Raise tooz error when unexpected last entries found
• etcd: properly block when using ‘wait’
• Share _get_random_uuid() among all tests
• Updated from global requirements
• Clean leave group hooks when unwatching
• Fix the test test_unwatch_elected_as_leader
• Updated from global requirements

1.41.0¶

• File driver: properly handle Windows paths
• Updated from global requirements
• Updated from global requirements

1.40.0¶

• Add docs for new consul driver
• Change dependency to use flavors
• Run doc8 only in pep8 target
• Move pep8 requirements in their own target
• zookeeper: do not hard depend on eventlet
• Remove unused iso8601 dependency
• tests: remove testscenario usage
• file: set no timeout by default
• tests: move bad_url from scenarios to static test
• Expose timeout capabilities and use them for tests
• Use pifpaf to setup daemons

1.39.0¶

• Updated from global requirements

1.38.0¶

• Using LOG.warning instead of LOG.warn
• Updated from global requirements
• redis: do not force LuaLock
• Fix coordinator typo
• Updated from global requirements
• Ensure etcd is in developer and driver docs

1.37.0¶

• Remove unused consul future result
• Updated from global requirements
• Add a consul based driver
• file: make python2 payload readable from python3

1.36.0¶

• Updated from global requirements

1.35.0¶

• Drop babel as requirement since its not used
• Updated from global requirements
• Updated from global requirements
• Updated from global requirements
• Updated from global requirements
• coordination: expose a heartbeat loop method

1.34.0¶

1.33.0¶

• Updated from global requirements
• Compute requires_beating
• Fix calling acquire(blocking=False) twice leads to a deadlock

1.32.0¶

• Raises proper error when unwatching a group

1.31.0¶

• Updated from global requirements
• Updated from global requirements
• Add .tox, *.pyo and *.egg to .gitignore
• Enable OS_LOG_CAPTURE so that logs can be seen (on error)

1.30.0¶

• Updated from global requirements
• Add lock breaking
• pgsql: fix hostname parsing
• Updated from global requirements
• Updated from global requirements
• Update voluptuous requirement
• Updated from global requirements
• Updated from global requirements
• Have zookeeper heartbeat perform basic get
• Add desired characteristics strict subset validation
• Add base64 key encoder (and validations)
• Use voluptuous instead of jsonschema
• Add programatic introspection of drivers characteristic(s)
• Updated from global requirements
• pep8: fix remaining errors and enable all checks
• Use utils.convert_blocking to convert blocking argument
• Adjust some of the zookeeper exception message
• Fix etcd env setup
• tests: do not hardcode /tmp
• utils: replace exception_message by exception_to_unicode
• Add a default port and default host
• etcd: driver with lock support
• Use utils.to_binary instead of using redis module equivalent
• Remove tested under 2.6 from docs

1.29.0¶

• Updated from global requirements
• Add basic file content schema validation
• Spice up the driver summary/info page
• Make all locks operate the same when fetched from different coordinators
• Add noted driver weaknesses onto the drivers docs
• Updated from global requirements
• File: read member id from file with suffix “.raw”
• Reduce duplication of code in handling multi-type blocking argument
• Updated from global requirements
• Add comment in memcache explaining the current situation with lock release

1.28.0¶

• Add ‘requires_beating’ property to coordination driver
• {pg,my}sql: fix AttributeError on connection failure
• tests: allow ipc to bypass blocking=False test
• pgsql: remove unused left-over code
• Add ‘is_still_owner’ lock test function

1.27.0¶

• Updated from global requirements
• Updated from global requirements
• Remove python 2.6 and cleanup tox.ini

1.26.0¶

• Updated from global requirements
• Allow specifying a kazoo async handler ‘kind’
• Updated from global requirements

1.25.0¶

• Updated from global requirements
• Add standard code coverage configuration file
• docs - Set pbr ‘warnerrors’ option for doc build
• Include changelog/history in docs
• Updated from global requirements
• Expose Znode Stats and Capabilities
• Allow more kazoo specific client options to be proxied through

1.24.0¶

• Updated from global requirements

1.23.0¶

• Changes to add driver list to the documentation
• Updated from global requirements

1.22.0¶

• Updated from global requirements
• Accept blocking argument in lock’s context manager
• Make RedisLock’s init consistent with other locks
• Updated from global requirements

1.21.0¶

• Raise exception on failed lock’s CM acquire fail
• Be more restrictive on the executors users can provide

1.20.0¶

• Updated from global requirements
• Updated from global requirements
• Use futurist to allow for executor providing and unifying
• Use a lua script(s) instead of transactions

1.19.0¶

• Updated from global requirements
• Change Lock.name to a property
• Update .gitignore
• Updated from global requirements
• Fixup dependencies
• Expose started state of coordinator to external
• Updated from global requirements
• Updated from global requirements

1.18.0¶

• Remove tooz/openstack as it is empty and not used
• Fix sp ‘seonds’ -> ‘seconds’
• Ensure run_watchers called from mixin, not base class
• Updated from global requirements
• Update compatibility matrix due to file drivers new abilities

0.17.0¶

• No longer need kazoo lock custom retry code
• Ensure unwatch_elected_as_leader correctly clears hooks

0.16.0¶

• Updated from global requirements
• Updated from global requirements
• Ensure lock(s) acquire/release returns boolean values
• Expose ‘run_elect_coordinator’ and call it from ‘run_watchers’
• Share most of the `run_watchers` code via a common mixin
• Remove 2.6 classifier
• Remove file-driver special no-async abilities
• Delay interpolating the LOG string
• Use `encodeutils.exception_to_unicode` for exception -> string function
• Use the `excutils.raise_with_cause` after doing our type check
• Updated from global requirements
• Use the ‘driver_lock’ around read operations
• Updated from global requirements
• Switch badges from ‘pypip.in’ to ‘shields.io’
• Updated from global requirements
• Add watch file driver support
• Make the file driver more capable (with regard to groups)
• Ensure locks can not be created outside of the root file driver directory
• Updated from global requirements
• Use MySQL default port when not set explicitly
• Use fasteners library for interprocess locks
• Implement watch/unwatch elected_as_leader for redis driver
• Updated from global requirements
• Use lua locks instead of pipeline locks
• Move more string constants to class constants with docstrings
• Updated from global requirements
• Updated from global requirements
• Remove support for redis < 2.6.0
• Expose Zookeeper client class constants
• Expose redis client class constants
• Use a serialization/deserialization specific exception
• Expose memcache coord. class constants
• Explicitly start and execute most transactions
• Provide and use a options collapsing function
• Add zookeeper tag in setup.cfg
• Use pymemcache pooled client
• Use oslo.serialization msgpackutils
• Provide ability for namespace customization for Zookeeper and Zake drivers
• Typo in Locking doc
• Move optional driver requirements to test-requirements.txt
• Have run_watchers take a timeout and respect it
• Heartbeat on acquired locks copy
• Avoid using a thread local token storage

0.15.0¶

• Fix param name to be its right name
• Replace more instance(s) of exception chaining with helper
• Just use staticmethod functions to create _dumps/_loads
• Uncap library requirements for liberty
• Link AOF to redis persistence docs
• Add exception docs to developer docs
• Add + use helper to raise + chain exceptions
• Allow the acquired file to be closed manually
• Updated from global requirements
• Silence logs + errors when stopping and group membership lost
• Make and use a thread safe pymemcache client subclass
• Handle errors that come out of pymemcache better
• Use rst inline code structure + link to sentinel

0.14.0¶

• Beef up the docstrings on the various drivers
• fix lock concurrency issues with certain drivers
• Add pypi download + version badges
• Denote that 2.6 testing is still happening
• Updated from global requirements
• Use a sentinel connection pool to manage failover
• fix mysql driver url parsing

0.13.1¶

• Switch to non-namespaced module imports
• Add a driver feature compatibility matrix
• Remove support for 3.3

0.13.0¶

• Two locks acquired from one coord must works
• Updated from global requirements
• Releases locks in tests
• Allow coordinator non-string options and use them
• Since we use msgpack this can be more than a str
• Updated from global requirements
• Avoid re-using the same timeout for further watcher ops

0.12¶

• retry: fix decorator
• file: fix typo in errno.EACCES

0.11¶

• Add a file based driver
• Upgrade to hacking 0.10
• Update sentinel support to allow multiple sentinel hosts
• Allow to pass arguments to retry()
• IPC simplification

0.10¶

• Add support for an optional redis-sentinel
• README.rst tweaks
• A few more documentation tweaks
• Sync requirements to global requirements
• Add create/join/leave group support in IPC driver
• Add driver autogenerated docs
• Update links + python version supported
• zookeeper: add support for delete group
• redis: add support for group deletion
• tests: minor code simplification
• memcached: add support for group deletion
• memcached: add support for _destroy_group
• Switch to using oslosphinx
• Add doc on how transaction is itself retrying internally
• Fix .gitreview after rename/transfer
• tests: use scenarios attributes for timeout capability
• tests: check for leave group events on dead members cleanup
• memcached: delete stale/dead group members on get_members()
• tests: remove check_port
• tests: do not skip test on connection error

0.9¶

• doc: add missing new drivers
• doc: switch examples to Zake
• doc: add locking
• Fix tox envlist
• Drop Python 3.3 tests in tox
• Allow tox with py34 and MySQL
• Test connection error scenarios on more drivers
• Translate psycopg2 errors/exceptions into tooz exceptions
• Ensure ‘leave_group’ result gotten before further work
• watch_leave_group not triggering callback on expired members
• Add MySQL driver
• Discard ‘self’ from ‘_joined_groups’ if we got booted out
• Implement non-blocking locks with PostgreSQL
• More retry code out of memcached
• Add a PostgreSQL driver
• Fix gate
• Handle when a group used to exist but no longer does
• tox: split redis/memcached env
• Fix memcached heartbeat on start()
• tox: splits test scenarios
• Add a minimum redis version check while starting
• Make requirement on redis 2.10.x explicit
• Try to use PSETEX when possible
• Use hdel with many keys where supported
• Avoid logging warnings when group deleted or member gone
• Ensure that we correctly expire (and cleanup) redis members
• Various fixes for locks and version compatibility
• Move sysv_ipc deps to test-requirements

0.8¶

• test: try to stop() first
• Convert the rest of memcached driver functions to futures
• Add a assertRaisesAny helper method
• Allow zake to be tested
• Add a redis driver
• Ensure groups leaving returns are gotten
• Raise the new OperationTimedOut when futures don’t finish
• Start to add a catch and reraise of timed out exceptions
• Adjust the timeout to reflect the repeated retries
• ipc: do not delete the lock if we never acquired it
• Add home-page field

0.7¶

• Split up the requirements for py2.x and py3.x
• ipc: Fix acquire lock loop logic

0.6¶

• Make lock blocking with no time out by default

0.5¶

• coordination: remove destroy() from the lock protocol
• IPC: fix a potential race condition at init
• Fix IPC driver on OS X
• Switch to oslo.utils
• Blacklist retrying 1.3.0
• Use futures to make parts of the memcached driver async
• Have examples run in the py27 environment and make them work

0.4¶

• Standardize the async result subclasses
• Fix the comment which was borrowed from the IPC driver
• Be more tolerant of unicode exceptions
• Standardize on the same lock acquire method definition
• Standardize on hiding the lock implementation
• On lock removal validate that they key was actually deleted
• Use a thread safe deque instead of a queue
• Change inline docs about class fake storage variable
• LOG a warning if the heartbeat can not be validated
• Add doc8 to the py27 test running
• Use the more reliable sysv_ipc instead of posix_ipc+lockutils
• Only start zookeeper/memcached when not already running
• Let zake act as a in-memory fully functional driver
• Switch to a custom NotImplemented error
• Ensure lock list isn’t mutated while iterating
• Move Zake driver code to separated Python module
• Work toward Python 3.4 support and testing
• Unlock the kazoo version
• Bump up zake to be using the newer 0.1 or greater
• Fix zake driver with latest release
• memcached: switch leader election implementation to a lock
• Add the generation of the documentation in tox.ini
• Add coverage report

0.3¶

• Switch to URL for loading backends
• Import network_utils from Oslo
• coordination: add IPC driver
• coordination: raise NotImplementedError as default
• Add documentation
• Upgrade hacking requirement
• memcached: use retrying rather than sleeping
• Use retrying instead of our custom code
• Update requirements file matching global requ

0.2¶

• memcached: implement leader election
• Fix a race condition in one of the test

0.1¶

• memcached: add locking
• coordination: implement lock mechanism in ZK
• coordination, zookeeper: add get_leader()
• coordination, zookeeper: implement leader election
• coordination: remove wrong comment in tests
• memcached: add support for leave events
• memcached: implement {un,}watch_join_group()
• coordination: raise GroupNotCreated when watching uncreated group
• coordination, zookeeper: add {un,}watch_leave_group
• coordination, zookeeper: add watch_join_group
• tests: skip test if function is not implemented
• coordination: add hooks system
• Add memcached driver
• zookeeper: use bytes as input/output type
• tests: test client disconnection
• coordination: add heartbeat method
• Add pbr generated and testr files to gitignore
• coordination: enhance MemberAlreadyExist exception
• coordination: enhance GroupNotCreated exception
• coordination: enhance MemberNotJoined
• coordination: enhance GroupAlreadyExist exception
• tests: test capabilities on non existent group/member
• tests: add a test for group already existing
• tests: fix variable name
• Fix the default prototype for join_group
• Adds basic tests which deals with exceptions
• Fixes TypeError in _leave_group_handler
• Remove _wrap_call_kazoo
• Add asynchronous API
• Delete models.py and clean get_members()
• Add a fake ZooKeeper driver
• Allow passing in a handler
• First commit of Tooz
• Added .gitreview