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 5.7.36 (release notes, download) and MySQL-wsrep 8.0.27 (release notes, download) with Galera replication library 3.36 (release notes, download) implementing wsrep API version 25 and Galera replication library 4.11 (release notes, download) implementing wsrep API version 26. This release incorporates all changes to MySQL 5.7.36 and 8.0.27 respectively, adding a synchronous option for your MySQL High Availability solutions.
For the Galera replication library 3.36, we ensure that join messages are always sent, even when the node is not int he JOINER state; this relaxation allows for the correct handling of the desync counter, when a node is in the DONOR state.
For the Galera replication library 4.11, we introduced progress reporting for Incremental State Transfers (IST) via the event service interface (see below for end user implementations); and now the GCache can also report ring buffer recovery progress via the event service interface. The ring buffer has also been enhanced to avoid segmentation faults if an invalid buffer leaks past the initial recovery scan. For applications that do not track the provider side transaction lifetime accurately, we added a defensive mechanism to not attempt releasing the transaction for writesets. We made some improvements around CMake to make compilation more efficient. It also includes all changes in 3.36.
In MySQL 5.7.36, we made some parallel apply improvements: CREATE TABLE..LIKE is now marked as unsafe and a transaction is marked as unsafe when it traverses into a cascade delete operation. (Fixes MDEV-26803.) CREATE TABLE with FOREIGN KEY constraint failed to apply in parallel, so now will get reference tables appended as keys in the writeset. (Fixes MDEV-27276.)
In MySQL 8.0.27, we have like MySQL 5.7.36 resolved the above. We have also added progress reporting for SST scripts, which now reports the SST stage and the estimated total work, and the completed work in the current stage. Progress reporting also works when you use CLONE SST. You are recommended to have the pv tool (while helps monitor the progress of data through a pipe).
We have also included a wsrep_recover utility script to simplify your cluster recovery procedures, that will determine the last committed GTID of a crashed or stopped node. We have also improved the xtrabackup-v2 script to work with the current XtraBackup 8.0.27. We have also added a new variable, innodb_wsrep_applier_lock_wait_timeout to configure the timeout in seconds after which the wsrep watchdog starts killing local transactions that are blocking the applier.
We have stopped building CentOS 8 packages, and we cannot guarantee that the packages built for Red Hat Enterprise Linux 8 will continue working for CentOS Stream.