v0.22.0

Features

• Bumped Percona XtraDB Cluster to v5.7.28-31 [#170064631]

Known Issues

• download-logs reports that fetching MySQL InnoDB Status logs failed, when in reality, it actually succeeded.
  This is a known defect - its because the download-logs tries to download InnoDB Status logs from the jumpbox VM,
  which doesn't exist. It is safe to ignore this error when using the download-logs script.

v0.21.0

Features

• Bumped Percona XtraDB Cluster to v5.7.27
  [#168584227]

• Bumped golang to 1.13
  [#168584227]

• Toggled default for relay_log_recovery. More context is available in the commit.

• download-logs fetches INNODB ENGINE STATUS as well.

Known Issues

• download-logs reports that fetching MySQL InnoDB Status logs failed, when in reality, it actually succeeded. This is a known defect - its because the download-logs tries to download InnoDB Status logs from the jumpbox VM, which don't exist. It is safe to ignore this error when using the download-logs script.

v0.20.0

Features

• Bumped percona-xtrabackup to v2.4.12
  [#165302029]

Bug Fixes

• Stale pid files are cleaned up so that processes start reliably.
  [#167387692]

• Checking the correct Galera property for cluster readiness so that redeploying database works reliably.
  [#167362064]

v0.19.0

Features

• BOSH DNS now checks proxy health status before forwarding traffic to a given node in HA cluster
  [#165867468]

Bug Fixes

• Set sync_binlog property such that each transaction is committed to disk except when running HA cluster
  [#166156735]

v0.18.0

Bug Fixes

• [BUG] Fixed the race condition between pxc-mysql and bosh-dns jobs during pre-start
  [#166173640]

v0.17.0

Certificate-based Client Authentication

• [FEATURE] You can now use certificate-based authentication to a MySQL instance from pxc-release.
  [#165911554]

Bug Fixes

• [BUG] The drain script no longer times out when inactive_mysql_port is configured.
  [#165627424]

Performance Improvements

• [FEATURE] We have exposed the wsrep_slave_threads Galera property so that it can be configured to prevent large write-sets from blocking Galera replication. Defaults to 1. [#165362955]
• [FEATURE] Removed proxy.max_open_files property, in favor of using the OS's max number of file descriptors. [#165362955]

Dependency Updates

• pxc-release uses Golang 1.12.x
  [#165268032]

Changes to Manifest

• pxc-mysql.engine_config.galera.wsrep_slave_threads
  • Defines the number of threads to use when applying replicated write-sets.
  • default: 1
• REMOVED proxy.max_open_files
  • Configure this number to be twice as large as mysql.max_connections
  • default: 3000

v0.16.0

BPM

We've made significant progress in moving to BPM (BOSH Process Manager). This
is a significant departure from how jobs are launched. Our goal is to
standardize on a set of templates, in contrast to the hand-written versions of
previous releases.

#161428228, #161428212, #162812889, #161428230

Features

Added the ability to create loosely coupled leader and follower nodes for
replication by configuring properties a set of properties. The full list of
configurable properties will be published soon.

#164024204, #164024349, #164156850, #164024217

Improvements

We’ve made improvements to better handle empty data directory on
initialization. #164185461

MySQL process starts only after successful execution of BOSH wait scripts.
#164843514

Added get-sequence-number utility, which is useful in determining health of HA
clusters #162070459

Bug Fixes

[BUG] Fixed a bug that started the MySQL process every time the data directory
was removed or reinitialized, which is a common support operation. #165186930

v0.15.0

BPM

We've made significant progress in moving to BPM (BOSH Process Manager). This is a significant departure from how jobs are launched. Our goal is to standardize on a set of templates, in contrast to the hand-written versions of previous releases. [#161428229], [#161428222]
[#161428240], [#161428221]

Bug Fixes

• Bug fix: There was a bug in the Xenial stemcell that prevented proper configuration of ulimit. This was fixed by moving jobs to BPM. [#162247238]
• Bug fix: Fixed issue where MySQL GRA logs were not being purged correctly [#163441056]

Performance Improvements

• Bug fix: There was a performance regression in Percona XtraDB Cluster in the SHOW GLOBAL VARIABLES query, which caused timeouts in proxy health checks. This was fixed by upgrading to Percona XtraDB Cluster v5.7.25-31.35 [#163595933]
• pxc-release uses all available cores during compilation to reduce time to compile [#161100650]

Security Improvements

• There are updates to the add-backup.yml and pxc.yml ops files in mysql-backup-release so that MySQL backups are run as a limited-permissions user, rather than as admin [#157238354]

Improvements to Migrating from cf-mysql-release

• We have better error messaging around improper migrations from cf-mysql-release to pxc-release, such as using incorrect ops-files [#159998828]

Dependency Updates

• pxc-release uses Golang v1.11.x [#160776067]
• pxc-release uses Percona XtraDB Cluster v5.7.25-31.35 [#164316826]

v0.14.0

v0.14.0

Features

• The drain script checks the health of the nodes in the cluster before draining to make sure that a node isn't drained if it is the only healthy node in the cluster. [#157656403]

Bug Fixes

• Bug fix: Drain script now kills the mysql process in a way such that connections are severed, so connections don't stay when the node leaves the cluster. [#157299144]

Dependency Updates

• Removed jquery dependency in the switchboard dashboard. [#157063299]

v0.13.0

v0.13.0

Bug Fixes

• Bug fix: Typo in engine_config.audit_logs.audit_log_exclude_accounts BOSH property usage in my.cnf template. [#159243291]