1.7. Tungsten Clustering 5.4.0 NYR (NYR)

Version End of Life. NYR

This is a bugfix release.

Behavior Changes

The following changes have been made to Continuent Tungsten and may affect existing scripts and integration tools. Any scripts or environment which make use of these tools should check and update for the new configuration:

  • The Connector passThroughMode configuration option is now deprecated.

    The following passThroughMode entry will be removed from tungsten-connector/conf/connector.properties. There is currently no tpm option for this, and it is undocumented. The default will be kept to passThroughMode=true.

    # The Tungsten Connector offers an extra fast data transfer mode known as
    # pass-through. When the following switch enabled (default), the Connector
    # will directly transfer data packets between the client and the server.
    # When disabled, every native MySQL command will be translated into a JDBC call.
    passThroughMode=true

    Issues: CT-897

Known Issue

The following issues may affect the operation of Continuent Tungsten and should be taken into account when deploying or updating to this release.

  • Fall-Back Bridge Mode is NOT compatible with MySQL 8 at this time.

    MySQL 8 support was added in versions 5.4.0 and 6.1.0, which broke the Fall-Back Bridge mode functionality. When switching to Bridge mode, the calling client is unable to complete the handshake due to authentication flags, namely CLIENT_DEPRECATE_EOF and CLIENT_CONNECT_ATTRS.

    Issues: CT-967

Improvements, new features and functionality

  • Command-line Tools

    • A new utility script has been added to the release, tungsten_post_process, which assists with the graceful maintenance of the static cross-site replicator configuration files on disk.

      Issues: CT-761

      For more information, see The tungsten_post_process Command (in [Tungsten Clustering (for MySQL) 6.0 Manual]).

    • A new utility script has been added to the release, tungsten_reset_manager, which assists with the graceful reset of the manager's dynamic state files on disk.

      Issues: CT-850

      For more information, see The tungsten_reset_manager Command (in [Tungsten Clustering (for MySQL) 6.0 Manual]).

Bug Fixes

  • Installation and Deployment

    • Fixing the rpm-based post-install chown command so that symlinked directories get correct ownership.

      Issues: CT-767

    • The Tungsten Clustering RPM now preserves the original OS group memberships for the tungsten user.

      Issues: CT-867

  • Command-line Tools

    • Long service names within cctrl could cause output to fail when displaying information. The underlying issue has been fixed. Because long service names can cause formatting issues, a new option, --cctrl-column-width has been added which can be used to configure the minimum column width used to display information.

      Issues: CT-773, CT-926

  • Tungsten Connector

    • MySQL ping commands are now reconnected/retried upon "server gone away" error (Proxy mode ONLY).

      Issues: CT-863

    • Introduced a new tpm flag allowing for tuning Connector thread stack size, which can be required in particular cases where large requests are sent as text to a connector configured for automated read/write splitting (smartscale and direct reads).

      The setting is commented out by default, leaving the JVM use its own default, generally 1024.

      Setting tpm option connector-thread-stack-size={value in kb} will override this value.

      Note

      Please note that since the new size will be allocated for each incoming connection, increasing the thread stack size will affect the total runtime memory used by the connector instance

      Issues: CT-973

  • Tungsten Manager

    • Fixed an edge case where the master node and the coordinator node are the same, then the node was rebooted. The failover would not complete and throws an error.

      Issues: CT-479

    • Remove spurious warnings during composite switch or failover.

      Issues: CT-487

    • Fixed a case when get_replicator_roles and cctrl ‘ls -l’ didn't work if a replicator was stopped.

      When a replicator is not running insert the Replicator.HOST to the ReplicationNotification. It was wrongly inserted into the Replicator.DATASERVERHOST. This fixes the get_replicator_roles script. Also substituted hard-coded strings for their constant values.

      Issues: CT-760, CT-876