MySQL Cluster Manager 8.0 User Manual
Version 8.0.26 is the first General Availability release of the 8.0 series of MySQL Cluster Manager. It is suitable for use with MySQL NDB Cluster 8.0, 7.6, and 7.5.
Listed below are functionality changes and bug fixes for MySQL Cluster Manager 8.0.40 , in comparison to release 1.4.8.
Important Change: There are a number of important changes going from MySQL Cluster Manager 1.4 and 8.0; see What Is New in MySQL Cluster Manager 8.0 for details. Users of MySQL Cluster Manager 1.4 should make sure they understand those changes before they migrate to MySQL Cluster Manager 8.0.
Agent: Status detection for a management or data node has become more efficient, as mcmd no longer relies on reading the log of the node to determine its readiness. (Bug #31446224)
Client: The file hierarchy under the
collected-filesfolder created by the collect logs command has been reorganized. See the description for the collect logs command for details. (Bug #27921828)Client: The mcm client delivered in the stand-alone MySQL Cluster Manager package for Linux platforms (which does not include the NDB Cluster binaries) no longer has a dependence on a mysql client already installed on the system. (Bug #14032410)
Agent: The memory size returned by the show settings command using the
--hostinfooption was not always consistent. It was because the value was not rounded off to the nearest megabyte by a consistent method, and this patch corrects the issue. (Bug #32788118)Agent: mcmd quit unexpectedly when running the import config command for a 8.0.24 or 8.0.25 cluster with an ndb_mgmd node ran with the
--print-full-configoption. (Bug #32785972)Agent: An update process command failed when a node was started with the
--nowait-nodesoption. With this fix, the option is allowed, but a warning is issued saying the option is not preserved after the update. The same applies now to the--initial-startand--logbuffer-sizeoptions, for the relevant node types. (Bug #32288203)Agent: When MySQL Cluster Manager was upgraded from 1.4.2 to 1.4.8, if the mcmd user on the
mysqldnodes hadlocalhostas its host, connections cannot be established to the mysqld nodes after the upgrade. It was because the upgrade changed the host for the user account to127.0.0.1. With this patch, when connection fails for'user'@'127.0.0.1', mcmd attempts'user'@'localhost'. (Bug #32131361)Agent: After assigning explicitly each node to a node group, the cluster cannot be started if the node group IDs did not start from zero or were not consecutive numbers. This patch removes the restrictions. (Bug #31837504)
Agent: The
delete sitecommand failed with the complaint that the site contains clusters even when it did not. (Bug #31739439)Agent: Where some command caused a process to fail (
Error 7006), a warning from the log was displayed in place of the error message that explained the actual problem. (Bug #31496860)Agent: Sometimes, a node remained running even after the mcm client reported that it had been stopped successfully. It happened when mcmd experienced troubles with hostname resolutions, and this patch eliminated the unnecessary hostname lookups. (Bug #31399903)
Agent: An mcmd agent might consume a disproportionately large amount of the host's CPU power when there was a long log of highly repetitive messages that mcmd tried to examine. This patch limits the extent by which mcmd would go backward in time when examining the log messages, in order to avoid the problem. (Bug #31364594)
Agent: An mcmd agent quit unexpectedly during an
import configcommand when a mysqld query returned an error or if the query failed due to a connection problem. (Bug #31336743)Agent: A get command failed with the complaint that the host's package was not defined if, after a cluster was created, a package was added to a host outside of the cluster, and then the host executed the get command on the cluster. It was because the mcmd agent did not look up package information from the configuration store, and this patch makes it do that. (Bug #31293854)
Agent: On Windows platforms, an add hosts operation failed with a timeout at an attempt to add cloud hosts using computer names. (Bug #31095584)
Agent: The create site command failed when the name of the host for the mcmd agent was resolved to 127.0.1.1 by the operating system (which is the case in, for example, the default network setup for Ubuntu systems). (Bug #31012284)
Agent: When all hosts of a cluster were restarted and remove hosts was the last command executed before that, if hostname resolutions failed for all the hosts with all the agents after the restart, the site for the cluster would then disappear from the site list. This was due to all the agents removing their hosts from the site in that situation, each thinking its own host had been removed by other agents. With this fix, at recovery, an agent now creates a repository backup before removing itself form a site, in order to avoid the inadvertent loss of the site configuration metadata. Also, a test on hostname resolution on all site members is performed before an agent removes itself from a site, and if the test fails, the remove hosts command is not executed. (Bug #30889457)
Agent: A memory leak occurred for a restore cluster command run for an NDB 8.0 cluster when the mcmd agent failed to make a connection to a mysqld node. (Bug #30761243)
Client: Hostnames longer than 64 characters were not supported by MySQL Cluster Manager. With this patch, the limit is extended to 255 characters. (Bug #29459484)
When the MySQL Cluster Manager binaries for Red Hat Enterprise Linux/Oracle Enterprise Linux 7 was run on a Red Hat Enterprise Linux/Oracle Enterprise Linux 8 system, the operating system complained that the
libnsl.so.1library was missing. However, the library was not really required. With this fix, the complaint disappears. (Bug #30996729)