Changes in MySQL Cluster Manager 8.0.28 (2022-01-18, General Availability)

• Functionality Added or Changed

• Bugs Fixed

Functionality Added or Changed

• Agent: MySQL Cluster Manager now supports encrypted cluster backup. See descriptions for the backup cluster and restore cluster commands for details. (Bug #30733027)

• Agent: A new status value for show status --cluster has been introduced: the failed status indicates that while all processes in the cluster have exited, unlike stopped, some nodes did not exit cleanly. (Bug #30723251)

• Agent: Users can now set or override at the mcmd command line any agent options configurable with the agent configuration file (with the exception of mcmd_password) by specifying the option in the format of --section_name.option=value, where section_name is name of the section in the configuration file that the option belongs to (for example, --mcmd.bind_port=12345 and --logger.level=DEBUG). See mcmd, the MySQL Cluster Manager Agent for details. (Bug #29239804)

• Agent: In the list backups output for agent backups of a site (that is, for list backups --agent mysite), the Parts column has been replaced with the Files column, to show how many files each backup consists of. (Bug #27946094)

• Agent: The show status --operation commands always returned <no message> for the Description column. The column now displays more useful information (for example, a status or error message returned by the listed command) when it is available. (Bug #22607982)

• Client: The scenarios that caused the 7006 Process Error to be thrown are now covered by five different error codes:

  • 7006: A cluster process fails

  • 7030: An unexpected reply when connecting to a cluster process

  • 7032: An error occurs while starting a cluster utility process

  • 7033: A probe for an available port or socket (for connecting to a process to be started or just started) fails

  • 7034: A cluster process cannot be started

  (Bug #33563302)

• Client: A new command list warnings, has been introduced for checking important warnings issued across the whole site during the executions of MySQL Cluster Manager client commands. See the command description for details. (Bug #33378770)

• Client: The help command has been improved, with its output explaining itself better. (Bug #33321016)

• Client: A new --retry option has been added to the import config command for reimporting the cluster configuration from the config.ini file after the cluster is already running. See the description of the option for details. This option should only be used at the instruction of Oracle Support. (Bug #30836499)

• Client: Reporting on errors related to file reading or writing has been improved with a new error code (106 FILE_ERROR) and better error descriptions. (Bug #26246700)

• Client: A new command, delete backup, has been introduced for deleting a cluster backup. See the command description for details. (Bug #24295069, WL #14620)

• Client: The get command now has a new --all option, using which creates two more columns in the output of the command:

  • Type: Shows the expected type of the configuration attribute.

  • Restart: Shows the expected process restarts needed when changing the configuration attribute.

  See the description for the --all option for details. (Bug #23561548)

• Client: The error messages returned by MySQL Cluster Manager for failed utility processes (restores, upgrades, and so on) have now been improved: Output from the utility processes themselves, when available, are now returned, (Bug #23173980)

• Client: Error messages provided by the mcm client were sometimes not helpful with clusters having mysqld nodes, because mcm failed to extract relevant information from the mysqld error log. With this fix, mcm returns more helpful information under the same situations. (Bug #17282820)

Bugs Fixed

• Agent: A lot of logging that was intended for the note log level appeared in the log when --level was set to info (the default value). With this fix, the information only appears with the log severity level set to note or below. (Bug #33720754)

• Agent: The mcmd log contained duplicated entries of log excerpts from ndb_mgmd nodes. The duplication has now been eliminated. (Bug #33635312)

• Agent: In some situations (for example, when the agent was already hanging on some commands), a stop agents command did not return a reply to the client. It was because a locked mutex had prevented a proper shutdown for the agent, and this fix corrects the issue. (Bug #33633155)

• Agent: At the restart of mcmd, an error was reported in the log that the package path in use by the current cluster did not contain the MySQL NDB Cluster executables. The error was introduced by a change in release 8.0.26 that caused the wrong path to be checked, and it has now been rectified. (Bug #33620437)

• Agent: mcmd quit unexpectedly during an import config operation if, when querying a management node, the --configdir was pointing to an old and invalid configuration cache. With this fix, the error is avoided by not using the --configdir option for the query. (Bug #33541431)

• Agent: After a delete site command failed due to issues with data repository access, a subsequent agent command made mcmd quit unexpectedly. With this fix, preliminary checks reject the delete site command if needed, instead of allowing an unsuccessful run. Also, reporting on repository file access issues has been improved. (Bug #30498763)

• Agent: When using the set command to specify a value for the mysqld system variable innodb_directories in quotes, the quotes were skipped when the value was written to the mysqld node's my.cnf file, which might cause any semicolons (;) in the value to be misinterpreted as special characters on some platforms. (Bug #29317574)

• Agent: When backing up a cluster, the operation finished without any warnings even if the backup for a data node or mysqld node was missing at the end. With this fix, a warning was issued in the situation, and the backup operation reports the problem. (Bug #27168861)

• Client: The restore cluster command returned success even if its --skip-nodeid option included all the applicable nodes. With this fix, an error is thrown instead in the situation. (Bug #33608705)

• Client: A backup cluster command might fail with a timeout even if the backup was actually successful, when there was not much data to back up and the success message for the command arrived even before mcmd started waiting for it. With this fix, the success message is resent in the situation to avoid a timeout. (Bug #33437273)

• Client: After a backup cluster command was aborted, the next backup cluster operation failed with the message that it had been aborted unexpectedly . It was due to a confusion with the backup IDs involved, and this patch fixes the issue. (Bug #29239804)