Skip to content
This repository has been archived by the owner on Sep 21, 2022. It is now read-only.

Releases: cloudfoundry/cf-mysql-release

v36.19.0

27 Feb 22:24
Compare
Choose a tag to compare

v36.19.0

Improvements

  • Operators may now configure plans as paid and may configure currencies other
    than Usd

Bug fixes

  • gra-log-purger now handles purging a large number of GRA logs, where previously
    it may have failed. Operators experiencing this bug may suddenly see a large
    drop in disk usage.

Dependency Updates

  • cf-mysql-release now uses MariaDB to 10.1.38

v36.18.0

06 Feb 00:01
Compare
Choose a tag to compare

v36.18.0

Dependency Updates

  • [Smoke Tests] Update mysql2 gem to 0.5.2 for compatibility with latest Ruby buildpack

Bug fixes

  • [BUG] cf-mysql-release can now run successfully when the vm hostname includes the string "ipsec".

v36.17.0

26 Dec 22:18
Compare
Choose a tag to compare

v36.17.0

Improvements

Dependency Updates

  • Update rack to 1.6.11 in the broker [#161855910]
  • Update jquery in switchboard to 3.0.0 [#157063299]

Bug fixes

  • [BUG] Fix regression when automatically configuring ulimit on Xenial stemcells [#161936173]

v36.16.0

09 Nov 20:22
Compare
Choose a tag to compare

v36.16.0

Improvements

  • Xenial stemcell support
    • Ubuntu Trusty will be out of support next year. Fear not, cf-mysql-release will run fine on Xenial as well.
  • Allow configuration of innodb_default_row_format in cf-mysql-release [#159711795]
  • [BUG] Address hard to recover from drain script situation [#157299144]
    • Nodes now sever all connections before doing a full stop.
  • Compile pxc, cf-mysql and backups with all available cores [#161100650]
    • Our build scripts will gobble up all available CPUs on your compile instances. This helps a lot with compile times when deploying a new version of the release the first time.

Community Contributions

  • cloudfoundry/cf-mysql-release #209: Add cluster health check to drain script before shutting down local node [#157656403]
    • Allows a node to refuse to leave unless the cluster is healthy. Activate this via the property, cf_mysql.mysql.enable_drain_healthcheck
  • cloudfoundry/cf-mysql-release #212: Add cf_mysql.mysql.bootstrap_enabled property [#158648239]
    • Advanced feature: cf_mysql.mysql.bootstrap_enabled makes it possible to deploy Galera clusters in conjunction with multi-CPI, keeps nodes that are not the first to spin up from assuming they're the bootstrap node.

Dependency Updates

  • cf-mysql-release v36.16.0 should use MariaDB 10.1.36 [#160187830]
  • Bump Ruby in the broker [#161388453]
  • RubyGems 2.6.14 in cf-mysql 36.14.0-(2.2) affecting cf-mysql [#158373415]
  • Bump nokogiri from 1.8.4 to 1.8.5 [#161047659]
  • cf-mysql-release uses Golang 1.11.x [#160776073]
  • Update Loofah in the broker [#161728147]

Changes to Manifest

  • cf_mysql.mysql.enable_drain_healthcheck
    • When this is enabled, the --skip-drain flag is required in order to delete a deployment. Drain healthcheck cannot be enabled when using the arbitrator. Enabling this ensures mysql nodes will verify the health of all other nodes before draining.
    • default: false
  • cf_mysql.mysql.innodb_default_row_format
    • Optional, the default row format of innodb tables
    • default: compact
  • cf_mysql.mysql.bootstrap_enabled
    • Enables usage of bootstrap procedure
    • default: true

v36.15.0

12 Jul 17:51
Compare
Choose a tag to compare

v36.15.0

Dependency Updates

Bug Fixes

  • Bug fix: cf-mysql get-sequence-number does not work when grastate.dat file contains -1 [#158616668]

v36.14.0

11 May 04:51
Compare
Choose a tag to compare

v36.14.0

Dependency Updates and Bug Fixes

  • Bump golang to v1.9.6 [#157375681]
  • Bug fix: Enabling remote admin access, then disabling it fails to disable it [#157067582]
  • Bug fix: Changes introduced in v36.12.0 prevented the drain script from stopping mysql [#157135700]
  • Bug fix: Disabling the mysql job via the cf_mysql_enabled property would cause the drain script to fail [#157496349]
  • Bug fix: The download-logs script would fail if your bosh cli was greater than version 2 [#157329890]

Changes to Manifest

  • No changes

v36.13.0

01 May 19:21
Compare
Choose a tag to compare

v36.13.0

  • Bug fix: The Service Broker should not be returning a 422 when the supplied arbitrary parameters are invalid [#156640222]

Changes to Manifest

  • No changes

v36.12.0

04 Apr 05:57
Compare
Choose a tag to compare

v36.12.0

  • Improved: smoke-tests can run in a known org and space [#156236077]
    • Useful to enable running of smoke-tests when ASGs are used to limit access to the database
  • Experimental: cf-mysql-release should no longer write /etc/mysql/my.cnf [#156406871], [#156242509]
    • Enables the experimental colocation and migration to pxc release

Dependency Updates and Bug Fixes

  • cf-mysql-release shall use MariaDB 10.1.32 [#155006426]
    • Galera updated to 25.3.23

Changes to Manifest

  • cf_mysql_enabled
    • Defaults to true; false value will prevent mysql from starting; Used to support migrations to pxc release.
  • cf.smoke_tests.use_existing_space
    • Defaults to false; When true smoke tests will use the spaces specified by cf.smoke_tests.space rather than create a new space. Only applicable when cf.smoke_tests.use_existing_org is also true.
  • cf.smoke_tests.space
    • Only used when cf.smoke_tests.use_existing_space is set to true`

v36.11.0

17 Feb 02:39
Compare
Choose a tag to compare

v36.11.0

  • New feature: As a Space Developer, I'd like to be able to create read-only bindings to a Service Instance [#129885695], [#154575192]
  • Improved: The Operator would like pre-seeded users' passwords to be updated when changed in the manifest [#153529269]
  • Improved: Many small usability updates to download-logs, including now relying on bosh ssh rather than having manual SSH set up independently.
    • As an Operator, download-logs should leverage bosh CLI where possible so that I don't have to figure out ssh on each IaaS [#154609725]
    • As an Operator, I'd like to be able to distinguish the artifacts generated by download-logs [#154299373]
    • Bug fix: download_logs should not fail when a customer has other jobs installed that aren't readable by user vcap [#148156397]
    • As an Operator, it would be nice if download logs could contain the output of bosh vms and bosh instances [#154299563]
    • As a support team member, I'd like download-logs to include the sst log files [#154013852]

Dependency Updates and Bug Fixes

  • cf-mysql-release shall use MariaDB 10.1.30 [#154149018]

    • MariaDB 10.1.31 failed to build cleanly, we'll look into it and catch up in a coming release.
  • Bump cf-mysql-release to use xtrabackup 2.4.8 [#153101615]

  • Bug fix: Cluster nodes should not hang in pre-start after unclean shutdown of cluster [#153572034]

  • Bug fix: remove logging to syslog from cf-mysql-release [#153301323]

    We discovered that in switching over to syslog-release, we hadn't fully stopped independently sending logs to syslog. We've pared that back.

    • A known issue remains with routing-release, which may be addressed some time in the future. Track issue #97 for any updates.

Deprecation Notice: ARP Cache

When a MySQL server VM disappears, BOSH will recreate that VM with the same IP address. When that new VM does not share the same ethernet MAC address as the old one, switchboard will not be able to direct connections to the new VM.

BOSH has a feature to clear the old ethernet address from switchboard's cache, but this feature used to default to off. For this reason, switchboard contained a duplicate feature to ensure that it is able to reach the new VM.

Recently, we discovered that this code does not work when BOSH DNS is enabled. This bug has been fixed, see below. However, in doing, we've learned that the switchboard feature does not work on GCP, and possibly other IaaSes.

The BOSH feature is designed to work on any IaaS that BOSH supports. In a coming release, we will drop the duplicate switchboard feature.

We strongly suggest that you run all of your Directors with flush-arp enabled. If you do not have flush-arp enabled, you may encounter issues where switchboard is not able to reach a back end node, even after successful resurrection.

  • Bug fix: ARP Cache flushing broken in switchboard when BOSH DNS is enabled [does not fix GCP] [#154302130]

Changes to Manifest

  • No changes

v36.10.0

21 Nov 17:20
Compare
Choose a tag to compare

Syslog Changes

  • Sally can see that cf-mysql-release generates syslogs using a method that does not interfere with other software in the deployment [#152567983]

Breaking Change: We've cut independent calls to syslog from cf-mysql-release and now depend on syslog-release to do that work for us. If you are using cf-mysql-deployment, simply use the enable-syslog.yml operations file, and provide the following variables:

  • syslog_address
  • syslog_port
  • syslog_transport

Deploying with the enable-syslog.yml operations file will affect only the cf-mysql deployment. You don't need to use it if you already use syslog-release as a BOSH add-on.

Minor Features and Bug Fixes

Changed default

In #144637793, we introduced the new property, cf_mysql.mysql.enable_local_file and set the default to false. This is inconvenient, because it affects commonly used tools such as mysqlimport. The property is still there, but we've switched the default to true.

  • enable_local_file default overly restrictive [#148389277]

Other Changes

  • [PR #185] Configureable slow query threshold [#151987215]
    • Introduces a new property, cf_mysql.mysql.long_query_time, which you can set to the minumum number of seconds a query must run before showing up in the slow query log.
  • Update versions of ruby and gems used by broker (security) [#148614637]
    • Updates dependencies to address security concerns.
  • Bug fix: cloudfoundry/cf-mysql-release #184: Roadmin user doesn't have read privs for admin operations? [#151643206]
  • Bug fix: MySQL fails to start if the node once had IPSec installed but IPSec is no longer installed [#152740268]
  • Bug fix: Binky and other applications should not be able to write directly to a database that is in upgrade mode. [#151236116]

New Documentation

  • cloudfoundry/cf-mysql-release #192: We've added a document, backup-restore.md, that covers how to scale down and ease the restriction on the storage engine so that you can do a full restore without having issues with the MySQL system tables.

Manifest Changes

  • Removed: syslog_aggregator.address
  • Removed: syslog_aggregator.port
  • Removed: syslog_aggregator.transport