Stolon Versions Save

v0.17.0

New features

• Support PostgreSQL 13 (#831)
• Add configurable store timeouts (#765)

and many other changes.

A big Thank You to everybody who contributed to this release!

v0.16.0

New features

• k8s store: patch pod annotations instead of doing a full update (#751)
• Make proxyCheckInterval and proxyTimeout configurable (#756)
• Make DefaultSyncTimeout infinite (#752)
• Document SyncTimeout cluster spec option (#749)

and many other changes.

A big Thank You to everybody who contributed to this release!

v0.15.0

New features

• Support PostgreSQL 12 (#727)
• Added wal-g examples (#726)
• Stolonctl spec: don't show null clusterspec options (#736)
• Prevent stolonctl init with empty clusterspec file (#702)

Bug Fixes

• Do pg_rewind only against primary instance (#698)

and many other changes.

A big Thank You to everybody who contributed to this release!

v0.14.0

New features

• Add sentinel prometheus metrics (#656)
• Store only the last 2 postgres timeline histories to not exceed the max value size of the underlying store (#651)
• Add keeper prometheus metrics (#639)
• stolonctl status can output status in json format (#628)
• Enable all k8s client auth plugins (#625 #632)
• It's now possible to define the advertise address and port in stolon keeper (#581)
• A new stolonctl register command was added to set service discovery information about keepers to an external service (currently only consul) (#557)
• Ability to auto restart an instance when updating a postgres parameter that requries a restart (#568)
• Add stolon clusterdata read/write subcommands (#578)
• Enable verbose e progress logging for pg_basebackup (#665)
• Implement timeouts for kubernetes api calls (#666)

Bug Fixes

• Avoid unneeded postgres instance reloads (#558)
• Allow special characters in pg-su-username (#658)
• Fix failover process if a keeper has filesystem errors (#637)
• postgres: use go database/sql context functions (#635)
• Use go database/sql context functions (#635)
• Fix hanging sentinels (#659)

and many other changes.

Upgrades notes.

• The stolonctl clusterdata command has been split into two subcommands:
  1. stolonctl clusterdata read which will be used to read the current clusterdata.
  2. stolonctl clusterdata write which will be used to write the new clusterdata into the new store.

A big Thank You to everybody who contributed to this release:

• Anton Markelov (@strangeman)
• Arunvel Sriram (@arunvelsriram)
• Aswin Karthik (@aswinkarthik)
• Ben Wheatley (@benwh)
• David Eichin (@daMupfel)
• Dinesh B (@dineshba)
• Don Bowman (@donbowman)
• Harry Maclean (@hmac)
• Krishnaswamy Subramanian (@jskswamy)
• Lawrence Jones (@lawrencejones)
• Milyutin Maksim (@maksm90)
• Mosab Ibrahim (@mos3abof)
• Nicolas Juhel (@nabbar)
• Prabhu Jayakumar (@prabhu43)

New features

• Add a stolonctl command to force fail a keeper (#546)
• Overcome PostgreSQL synchronous replication limitation that could cause lost transactions under some events (#514)
• Users can now define archiveRecoverySettings in the cluster spec of a standby cluster. One of the possible use cases is to feed the standby cluster only with archived logs without streaming replication. (See Upgrade Notes) (#543)
• Keeper: remove trailing new lines from provided passwords (#548)

Bug Fixes

• Sort keepers addresses in pg_hba.conf to avoid unneeded postgres instance reloads (#558)
• Set recovery_target_action to promote when using recovery target settings #545)
• Fixed wrong listen address used in pg_hba.conf when SUReplAccessStrict mode was enabled (#520)

and many other bug fixes and documentation improvements.

Thanks to everybody who contributed to this release.

Upgrades notes.

• The clusterspec standbySettings option as been replaced by the standbyConfig option. Internally it can contain two fields standbySettings and archiveRecoverySettings (see the clusterspec doc with the descriptors of this new option). If you're updating a standby cluster, BEFORE starting it you should update, using stolonctl, the clusterspec with the new standbyConfig option.

New features

• Detect and report when keeper persistent data dir is not the expected one (usually due to wrong configuration, non persistent storage etc...) (#510)
• Support PostgresSQL 11 (beta) (#513)
• Replication slots declared in the clusterspec additionalMasterReplicationSlots option will now be prefixed with the stolon_ string to let users be able to manually create/drop custom replication slots (See Upgrade Notes) (#531)

Bug Fixes

• fix wrong address in pg_hba.conf when clusterspec defaultSUReplAccessMode is strict (#520)

and many other bug fixes and documentation improvements.

Thanks to everybody who contributed to this release:

Alexandre Assouad, Lothar Gesslein, @nseyvet

Upgrades notes.

• Replication slots declared in the clusterspec additionalMasterReplicationSlots option will now be prefixed with the stolon_ string to let users be able to manually create/drop custom replication slots (they shouldn't start with stolon_). Users of these feature should upgrade all the references to these replication slots adding the stolon_ prefix.

New features

• In the k8s store backend, stolon components discovery now uses the component label instead of the app label (See Upgrade Notes) (#469)
• Improved docker swarm examples to resemble the k8s one (#482)
• If the user enabled ssl/tls use it also for replication/pg_rewind connections (#501)
• Remove final newline from example base64 password in k8s example (#505)

Bug Fixes

• Fixed wrong libkv store election path (See Upgrade Notes) (#479)
• Fixed a check in synchronous replication that will block future synchronous standbys updates under some circumstances (#494)
• Fixed atomic writes of postgresql genenerated files (#495)

Thanks to everybody who contributed to this release:

Bill Helgeson, Niklas Hambüchen, Sylvere Richard, Tyler Kellen

Upgrades notes.

• In the k8s store backend, the label that defines the kind of stolon component has changed from app to component. When upgrading you should update the various resource descriptors setting the k8s component name (stolon-keeper, stolon-sentinel, stolon-proxy) inside the component label instead of the app label.
• When using the etcdv2 store, due to a wrong leader election path introduced in the last release and now fixed, if your sentinel returns an election error like election loop error {"error": "102: Not a file ... you should stop all the sentinels and remove the wrong dir using etcdctl rmdir /stolon/cluster/$STOLONCLUSTER/sentinel-leader where $STOLONCLUSTER should be substituted with the stolon cluster name (remember to set ETCDCTL_API=2).

v0.10.0

New features

• Initial support for native kubernetes store (#433)
• Improved sync standby management (#444)
• Ability to use strict and dynamic hba entries for keeper replication (#412)
• Ability to define additional replication slots for external clients (#434)
• Improved wal level selection (#450)

Thanks to everybody who contributed to this release:

Pierre Alexandre Assouad, Arun Babu Neelicattu, Sergey Kim

v0.9.0

New features

• The logs will be colored only when on a tty or when --log-color is provided (#416)
• Now the store prefix is configurable --store-prefix (#425)

BugFixes

• Fixed keeper missing waits for instance ready (#418)
• Fixed etcdv3 store wrong get leader timeout causing stolonctl status errors (#426)

Thanks to everybody who contributed to this release:

Pierre Fersing, Dmitry Andreev

v0.8.0

New features

• Added support for etcd v3 api (using --store-backend etcdv3) (#393)
• Now the stolon-proxy has tcp keepalive enabled by default and provides options for tuning its behavior (#357)
• Added removekeeper command to stolonctl (#383)
• Added the ability to choose the authentication method for su and replication user (currently one of md5 or trust) (#380)

BugFixes

• Fixed and improved db startup logic to handle a different pg_ctl start behavior between postgres 9 and 10 (#401)
• Fixed keeper datadir locking (#405)

and many other bug fixes and documentation improvements.

Thanks to everybody who contributed to this release:

AmberBee, @emded, Pierre Fersing