dbdeployer 2.4.2 is a bug-fix release, focused on MariaDB and Galera/PXC deploys. Upgrading is recommended for anyone deploying MariaDB with 2.4.0 or 2.4.1.
MariaDB deploys are fast again
In 2.4.0 and 2.4.1, every MariaDB node without Galera waited about two minutes for a Galera readiness signal it would never send. A 3-node MariaDB replication sandbox took around 8 minutes; with 2.4.2 it takes about 20 seconds (#142).
Whether a node waits for Galera readiness is now decided when the sandbox is created: only Galera and PXC cluster nodes wait, and other sandboxes no longer query the server for it at all.
Galera and PXC: joining nodes are checked, failures are reported
- Joining Galera/PXC nodes are now actually waited on until they have joined the cluster. Previously the check on those nodes could not log in, gave up silently after 30 seconds, and let the deploy continue unchecked. Galera/PXC deploys are also faster as a result.
- Behavior change: a Galera or PXC node that has not joined the cluster within 120 seconds now fails the deploy with an error that names the node and its log file. Previously the deploy continued and failed later, at grants, with a less obvious error.
MariaDB 11.4, 12.3 and 13.0 verified
MariaDB 11.4.13, 12.3.3 and 13.0.2 deploy as single and replication sandboxes, using the MariaDB replication syntax (CHANGE MASTER TO). They are now part of CI, next to 10.11.9, so they stay working (#139, #140, #141). MariaDB 12.x and 13.x binaries need glibc 2.28 or newer.
MariaDB releases that are not in dbdeployer's tarball list can be downloaded by version:
dbdeployer downloads get-by-version 12.3.3 --flavor=mariadb --unpack
Upgrading
No configuration changes are needed. Existing sandboxes keep working; new deploys get the fixed behavior. Full details are in CHANGELOG.md.