Release Notes for MariaDB MaxScale 23.02.4

With Binlog Router (binlogrouter), if a large transaction results in the creation of a binlog file larger than 4GB, that binlog file is corrupt. (MXS-4691)

With Binlog Router (binlogrouter), large transactions are skipped, resulting in data inconsistency on a slave. (MXS-4677)

With the Avro Router (avrorouter), MaxScale (mxs1687_avro_ha) can crash when the replication event is off by two bytes. (MXS-4674)

with replication via Binlog Router (binlogrouter), when the semi-synchronous replication is not configured for the server from which the binlogrouter is replicating, the binlogrouter eventually stops working. (MXS-4668)

With Binlog Router (binlogrouter), if the transaction being replicated does not fit in memory, the process will be killed or a std::bad_alloc will be thrown. (MXS-4690)

With mariadbclient, possible hang when connection_init_sql_file queries are executed. (MXS-4719)

The maxscale.cnf default file references MaxScale 6 documentation and if you run REPLICATION CLIENT, FILE, SUPER, RELOAD, PROCESS, SHOW DATABASES, EVENT grants, you will get an error with current versions of MariaDB. (MXS-4708)

With a readconnroute service, when a connection pool is connected, the session command response verification unnecessarily stores prepared statements, which increases CPU usage for each prepared statement. (MXS-4726)

SELECTSELECT statements targeting the information_schema are cached and fail to invalidate appropriately, for example when a table is dropped. (MXS-4717)

• Starting with this release, SELECT statements that target the information_schema are not cached.

When the ALTER, DROP, and RENAME statements are performed on a table, cache fails to invalidate cached SELECTSELECT statements on the table that has been changed. (MXS-4706)

With MaxScale GUI, default value of inputs in the object creation dialog is empty after closing the dialog. (MXS-4730)

With Binlog Router (binlogrouter), GTID values are compared as 32-bit integers rather than 64-bit integers. (MXS-4700)

With Galera Monitor (galeramon), replication lag is not updated for replicating servers. (MXS-4721)

With MariaDB Monitor (mariadbmon), switchover can fail with an Unknown thread id error (1094). (MXS-4675)

With MariaDB Monitor (mariadbmon), GTID updates may block REST API traffic. (MXS-4701)

The query classifier does not correctly parse a RENAME statement. (MXS-4714)

With Read/Write Split Router (readwritesplit), interrupted queries are retried even if all backends have encountered a permanent failure. (MXS-4696)

• Starting with this release, retries are not performed if all backends have encountered a permanent failure, resulting in faster reporting of errors to the client.

Replication via Binlog Router (binlogrouter) temporarily blocks the REST API. (MXS-4685)

With the query classifier, SHOW TABLE STATUS FROM ... fails with Schema Router (schemarouter). (MXS-4704)