Changes in MySQL Cluster Manager 1.4.4 (2017-10-05)

This section documents all changes and bug fixes that have been applied in MySQL Cluster Manager 1.4.4 since the release of MySQL Cluster Manager version 1.4.3.

• Functionality Added or Changed

• Bugs Fixed

Functionality Added or Changed

• Agent: On Windows platforms, when the collect logs command was run and the resulting file path for any of the collected files was longer than 255 characters, the command failed with ERROR 101 (00MGR) Access denied. A new type of error, PATH_TOO_LONG (106) Path max length exceeded for file filename, has been added and is now returned at the situation. (Bug #26619163)

Bugs Fixed

• Agent: On Solaris platforms, the collect logs command created lots of zombie threads after it finished. It was because a number of thread joins were missing in the code, which have been added by this fix. (Bug #26665347)

• Agent: For a large NDB Cluster site, the collect logs command frequently failed due to errors with the network connections during file transfers. This was because some error codes related to the network connections were not properly handled, and this fix corrects that. (Bug #26642616)

• Agent: The import config command sometimes failed, as the mcmd agent updated its configuration store before all the values to be imported arrived. This fix makes mcmd wait for all the values before starting the updates. (Bug #26610874)

• Agent: The collect logs command reported success even when an agent failed and the file transfer from it was incomplete. With this fix, when connection to an agent fails, the command allows running file transfers to complete, and then throws an error. (Bug #26594948)

• Agent: The collect logs command sometimes failed with ERROR 1003 Internal error: No clients connected, even though clients were connecting and files were being transferred. That was due to a faulty mechanism for tracking client connections, which has been corrected by this fix. (Bug #26575768)

• Agent: The mcm client hung when running the collect logs command if the host names for the cluster nodes could not be resolved, even if the cluster was actually defined using IP addresses. (Bug #26564176)

• Agent: A start cluster command timed out when its timeout value, being scaled for handling a large DataMemory size for a cluster, was accidentally reset. This fix makes sure the timeout for the execution step of a start cluster command is never lowered while the data nodes are starting. (Bug #26543383)

• Agent: After a node had failed and an update process command on it also failed, a stop cluster or stop process command that followed ran into an assertion error. It was because the node, automatically restarted after its failure, had its PID file modified by the update process command that was still running. This fix adds a check on the status of the target process for the update process command to prevent the restarted node's PID file from being modified. (Bug #26388603)

• Agent: The collect logs command hung for 30 seconds after completing all file transfers before returning the success message. (Bug #26362910)

• Agent: When a management node's status went from starting to stopped, it was not reported by its mcmd agent, causing commands like delete cluster to fail because the management node was taken to be still running. (Bug #26288009)