Codership is pleased to announce a new Generally Available (GA) release of the multi-master Galera Cluster for MySQL 5.7 and 8.0, consisting of MySQL-wsrep 5.7.35 (release notes, download) and MySQL-wsrep 8.0.26 (release notes, download) with Galera replication library 3.35 (release notes, download) implementing wsrep API version 25 and Galera replication library 4.10 (release notes, download) implementing wsrep API version 26. This release incorporates all changes to MySQL 5.7.35 and 8.0.26 respectively, adding a synchronous option for your MySQL High Availability solutions.
For the Galera replication library 3.35, we have updated the OpenSSL library version to 1.1.11 in tarball releases. We have also decided to End of Life building for SUSE Enterprise Linux Server 12.
For the Galera replication library 4.10, we improved error handling during Incremental State Transfers (IST). Going forward, we are replacing phrases more in tune with upstream, such that the provider option gcs.fc_master_slave is now deprecated in favour of gcs.fc_single_primary. OpenSSL 1.1.11 is available now, as are packages for Debian 11 (Bullseye) are now provided (while we have deprecated SUSE Enterprise Linux Server 15).
In MySQL 5.7.35, we improved node performance by removing any stalls during cluster write conflicts when innodb_thread_concurrency was enabled, as well as when TRUNCATE TABLE is run anda query is in the cleanup phase. We have also added a new wsrep_mode for IGNORE_CASCADING_FK_DELETE_MISSING_ROW_ERROR to workaround apply errors during a cascading delete. This fixes MySQL#80821. We have added new State Snapshot Transfer (SST) options for donor-timeout and joiner-timeout. A node is able to execute FLUSH TABLES even when it is not part of the primary component. Naturally, we have also decided to End of Life building for SUSE Enterprise Linux Server 12.
In MySQL 8.0.26, we have like for 5.7.35 resolved MySQL#80821 by introducing a new wsrep_mode for IGNORE_CASCADING_FK_DELETE_MISSING_ROW_ERROR. In some instances with streaming replication use, a bug caused the transaction to roll back locally but not within the cluster, and this was fixed. This fixes MDEV-26707. We have improved applying writesets by introducing periodic checks for long lock waits for applied writesets.
We have deprecated the use of “master” and “slave”, and will continue doing this within the documentation and in anything else going forward. Examples in this case are moving wsrep_restart_slave to wsrep_restart_replica and wsrep_slave_threadsto wsrep_applier_threads. Please read the release notes for more information.
We have also provided Debian 11 (Bullseye) binaries now, and naturally also deprecated SUSE Enterprise Linux Server 15.