MySQL Cluster Manager 1.4 Release Notes  /  Changes in MySQL Cluster Manager 1.4.5 (2018-01-25)

Changes in MySQL Cluster Manager 1.4.5 (2018-01-25)

Functionality Added or Changed

  • Agent: When the rotate log command is run, the new file name for the old mcmd log now has the timestamp put right before the file extension and preceded by an underscore; for example, mcmd.log becomes mcmd_2017-07-15T22-45-28.log after log rotation. (Bug #26696293)

  • Client: When an Internal Error (Error 1003) is returned, the error message now indicates the host on which the error occurs. (Bug #27082475)

Bugs Fixed

  • Agent: The reported status of a node was sometimes ignored (for example, a node reported to have stopped was next considered running). This was because log sequence numbers (LSNs) on nodes (by which mcmd tells whether the status report is out-of-date) were wrongly assigned sometimes. The fix corrects the LSN mechanism, so that status reports can be properly handled. (Bug #27338269)

  • Agent: A stop cluster command timed out while waiting for a failed management node to respond. With this fix, the command errors out immediately in the situation. (Bug #27253476)

  • Agent: mcmd hung on a restore cluster command when it failed to get the process handle of a node that was newly spawned and then failed. This fix adds proper error handling for the situation. (Bug #27225001)

  • Agent: After bootstrapping an NDB Cluster with the start cluster --bootstrap command, connecting to the cluster with a mysql client using the command suggested by MySQL Cluster Manager failed, because the command referred to the local host by its host name, which could be unresolvable in some situations. With this fix, the IP address 127.0.0.1 is used instead in the command. (Bug #27169533)

  • Agent: An mcmd agent exited unexpectedly when a ndbmtd node failed during a rolling restart, because the failure disallowed mcmd to identify the node's mirror. With this fix, mcmd throws the proper error in the situation and then continues with its operations. (Bug #27105336)

  • Agent: A start cluster command failed when there was an unmanaged node running on one of the cluster hosts. (Bug #27060818)

  • Agent: A stopping of a single node sometimes caused a large number of repeated exit reports to be generated to all mcmd agents. (Bug #27049206)

  • Agent: mcmd reported success for a restore cluster command even if it had failed. (Bug #25117413)

  • Agent: On Windows platforms, mcmd gave an excessive number of warnings to its log when it Could not get process handle for a cluster node. (Bug #23570996)

  • Agent: When a stop cluster command was running, sometimes a data node was not shut down completely and the command failed with an ERROR 7006 for nodegroup failure of the data nodes. This fix stops the unnecessary error from being thrown in the situation and lets the command continue. (Bug #21508316)

  • Agent: The import cluster command failed when the cluster to be imported used connection pooling for its mysqld nodes. (Bug #19357267)

  • Agent: On Windows platforms, after a cluster reconfiguration, when a management node failed to be restarted, it was still reported by mcmd as starting. It was because on Windows platforms, mcmd continued its attempt to connect to the stopped ndb_mgmd node in the situation until timeout. With this fix, status of the ndb_mgmd nodes are first checked before attempts were made to connect to them. (Bug #14255038)

  • Agent: When one of the NDB Cluster's ndb_mgmd nodes was already down, using a stop cluster --force to shut down the whole cluster was going to take a long time. It was because the stop process command's timeout, after being triggered, was not made known to all relevant processes on the cluster. With this fix, all mcmd agents are now informed of the timeout, so they will just kill processes that are still running after the timeout period. (Bug #13778840)

  • Agent: If a data node was added but could not be started at first, nodegroup creation failed even after the node was started successfully at the end. (Bug #13261667)

  • Client: When the show status command reported that a node had failed, if the mcmd agent that reported the failure also failed afterwards and the user then tried to restart the node, an ERROR 7007 was returned, saying wrongly that the node was already running. With this fix, a proper Error 1003 is returned with the message Agent on host hostname is unavailable. (Bug #27082586)

  • Client: Using the --removedirs option with the delete cluster command did not remove a data node's file system (the ndb_nodeID_fs directory) if it was located outside of the node's data directory. (Bug #25591980)

  • Client: While the set command cannot be used to set attributes for an unmanaged node that does not appear in the config.ini file (for example, the --port option for an unmanaged mysqld node), such attempts were not rejected by the mcm client. With this fix, mcm throws an error in the situation. (Bug #25512477)

  • Client: In some situations, after a new process was added, the start process --added command failed. It was because the added process was misidentified with a process that had stopped earlier. With this fix, the handling of stopped and added processes has been improved to prevent such cases of misidentification. (Bug #25160942)

  • Client: A stop cluster command failed if all data nodes of the cluster were in the failed status when the command was issued. (Bug #24931681)