Release Notes for MariaDB Enterprise ColumnStore 22.08.7

Overview

MariaDB Enterprise ColumnStore 22.08.7 is a maintenance release of MariaDB Enterprise ColumnStore. MariaDB Enterprise ColumnStore is a columnar storage engine included with MariaDB Enterprise Server.

MariaDB Enterprise ColumnStore 22.08.7 was released on 2022-12-21. This release is of General Availability (GA) maturity.

MariaDB Enterprise ColumnStore 22.08 replaces MariaDB Enterprise ColumnStore 6 in MariaDB Enterprise Server 10.6.

This release of MariaDB Enterprise ColumnStore is included with MariaDB Enterprise Server 10.6.11-6.

Users of earlier MariaDB Enterprise ColumnStore 22.08 releases are encouraged to upgrade.

Issues Fixed

Can result in a hang or crash

  • When the ExeMgr facility in the PrimProc process is restarted, if the SELECT component of an UPDATE statement fails, the DMLProc process performs a ROLLBACK operation, which can hang if the ExeMgr facility is not yet available. (MCOL-5263)

    • Starting with this release, if this scenario occurs, DMLProc avoids the hang by retrying the operation after a timeout period.

  • When ExeMgr is writing to the network, the process can sporadically crash due to a race condition. (MCOL-5264)

Can result in unexpected behavior

  • When the primary node is restarted, CMAPI and MaxScale sometimes choose different nodes to be the primary node. (MCOL-5293, MCOL-5306)

    • In previous releases, after the primary node was restarted, CMAPI would continue using the old primary server, but MaxScale would failover to a new primary server. If a user tried to select data from the new primary node, the operation could fail with the following error until the node is restarted:

      ERROR 1815 (HY000): Internal error: DBRM is not responding. Cannot accept queries
      
    • Additionally, the new primary node could write messages like the following to ColumnStore's debug.log file:

      NODE_NAME controllernode[409]: 58.561877 |0|0|0| E 29 CAL0000: DBRM: error: SessionManager::getSystemState() failed (network)   %%10%%
      NODE_NAME messagequeue[409]: 59.588469 |0|0|0| E 31 CAL0000: messageqcpp::hostnameResolver Name or service not known            %%10%%
      NODE_NAME messagequeue[409]: 01.642554 |0|0|0| E 31 CAL0000: messageqcpp::hostnameResolver Name or service not known            %%10%%
      
    • Starting with this release, after the primary node is restarted, both CMAPI and MaxScale should failover to the same primary server.

  • When a CTE contains a query that joins an InnoDB table and a ColumnStore table, if the CTE results contain a TIMESTAMP column, the column can sometimes be returned as a zero-date (which is 0000-00-00 00:00:00). (MCOL-5311)

  • When a VARCHAR column is NULL and is cast to INT, the conversion would fail, and PrimProc would excessively write log messages to its systemd journal. (MCOL-5346)

    • In previous releases, the systemd journal for PrimProc could contain many messages like the following:

      error in int conversion from ''
      
    • Starting with this release, PrimProc no longer writes those log messages to its systemd journal.

Platforms

In alignment to the enterprise lifecycle, MariaDB Enterprise ColumnStore 22.08.7 is provided for:

  • CentOS 7 (x86_64)

  • Debian 11 (x86_64, ARM64)

  • Red Hat Enterprise Linux 7 (x86_64)

  • Red Hat Enterprise Linux 8 (x86_64, ARM64)

  • Red Hat Enterprise Linux 9 (x86_64, ARM64)

  • Rocky Linux 8 (x86_64, ARM64)

  • Rocky Linux 9 (x86_64, ARM64)

  • Ubuntu 20.04 (x86_64, ARM64)

  • Ubuntu 22.04 (x86_64, ARM64)