Changes in MySQL Cluster Manager 1.4.0 (2015-12-07)

This section documents all changes and bug fixes that have been applied in MySQL Cluster Manager 1.4.0 since the release of MySQL Cluster Manager version 1.3.6.

• Functionality Added or Changed

• Bugs Fixed

Functionality Added or Changed

• Packaging: MySQL Cluster Manager is now built and shipped with GLib-2.44.0, OpenSSL 1.0.1p, and the MySQL 5.6 client library. (Bug #22202878)

• Agent: A warning is now logged (if log-level=warning) when a failed process is not restarted because the parameter StopOnError is set to true. (Bug #21575241)

• Agent: When using the import cluster command, if a mysqld node was started on the command line with options outside of a special, pre-defined set, the import failed with the complaint that those options were unsupported. Now, the import continues, as long as those options and their values are also included in the node's configuration created by MySQL Cluster Manager for import. (Bug #21363732)

• Agent: Two new options have been introduced for the upgrade cluster command: --retry and --nodeid. They, together with the --force option, allow a retry after an initial attempt to upgrade a cluster has failed. See the description for upgrade cluster for details. (Bug #20469067, Bug #16932006, Bug #21200698)

• Client: The get command now returns attributes in the same order as the MySQL Cluster ndb_mgmd command does when the --print-full-config option is used, with the non-data nodes being listed first and other nodes listed in increasing order of their node IDs. (Bug #22202973)

• Client: A new autotune command has been introduced, which tunes a number of parameters of the cluster to optimize its performance. (Bug #22202855)

• Client: The show settings command has a new --hostinfo option, with which the command prints out information on the host the mcm client is connected to. (Bug #21923561)

• Client: You can now use the wildcard * (asterisk character) to match attribute names in a get command. See The get Command for examples. (Bug #18069656)

Bugs Fixed

• Agent: On Windows platform, after a cluster import, the subsequent cluster restart timed out if a non-default value of the option --pid-file had been imported for a mysqld node. (Bug #21945519)

  References: This issue is a regression of: Bug #21111944.

• Agent: When a data node could not be restarted after a set command because some attributes were set wrongly, another set command could not be used to correct the attributes, because the set command required the data node to be running. With this fix, the second set command can now be executed even when the data node is not running, as long as the --force option is used. The failed node is then restarted, followed by a rolling restart of the cluster. (Bug #21943518)

• Agent: restore cluster timed out when the number of tables in the cluster was huge (>1000). It was because a timeout extension was blocked. This fix unblocks the extension. (Bug #21393857)

• Agent: At the initial startup of a large cluster (with memory size on the order of 10GB), the process might time out while waiting for a data node to start. This fix makes the transaction timeout longer for data node initiation. (Bug #21355383)

• Agent: Under some conditions, a show status command might report negative node group ID values for processes after an add process command was completed. That was because the agent reported the node group IDs before their proper values had arrived, after the creation of new node groups. This fix makes the agent wait for the correct node group IDs before reporting them. (Bug #21346804)

• Agent: After successful the execution of an add process and a subsequent start process --added command, a third command that was issued very shortly afterward might fail. This was due to the way the updates for the processes' statuses were handled after the new nodes were added, which has now been corrected. (Bug #21138604)

  References: See also: Bug #21346804.

• Agent: Setting a value for a “key-only” option for a MySQL node (that is, an option that does not take a value—for example, skip_show_database) with the set command and restarting the cluster afterward caused mcmd to attempt a cluster upgrade and back up the cluster. (Bug #21098403)

• Agent: The create site command sometimes failed with the error message “Lost connection to MySQL server during query.” It was due to an error in the code that handled the socket, which has now been fixed. (Bug #21027818)

• Agent: Parameters listed under the [mysqld default] or [tcp default] section of the config.ini file were not imported as configuration parameters for unmanaged API nodes. (Bug #20889471)

• Client: Output of the get command used with the --include-defaults (-d) option did not include matching TCP attributes that had default values. (Bug #21895322)