Continuent Documentation

Page in Other Manuals

8.7. The cluster_backup Command

The cluster_backup command provides a simple mechanism to execute a Tungsten Replicator backup inside of the cluster. It is designed to be called manually or as part of cron. The command should be added to cron on every server. When started, the command will check if the server is the current coordinator for the cluster. If not, the command will exit without an error. This design ensures that the command will only run on one server in the cluster.

The command supports command-line options that allow you to alter how and where the backup is executed.

cluster_backup [ --agent ] [ --datasource ] [ --directory ] [ --help, -h ] [ --info, -i ] [ --json ] [ --net-ssh-option ] [ --notice, -n ] [ --offline-backup String ] [ --quiet, -q ] [ --require-automatic-mode String ] [ --require-slave-backup String ] [ --validate ] [ --verbose, -v ]

Where:

Table 8.17. cluster_backup Command-line Options

OptionDescription
--agentThe replicator backup agent to use when running the backup, as configured in cctrl
--datasourceExecute the backup against the specified datasource
--directoryUse this installed Tungsten directory as the base for all operations
--help, -hDisplay this message
--info, -i
--jsonProvide return code and logging messages as a JSON object after the script finishes
--net-ssh-optionSet the Net::SSH option for remote system calls
--notice, -n
--offline-backup StringPut the datasource OFFLINE prior to taking the backup.
--quiet, -q
--require-automatic-mode StringThe script will fail if the cluster is not in the AUTOMATIC policy mode
--require-slave-backup StringRequire that the backup is taken of a slave datasource. If this is enabled and there are no slaves available, no backup will be taken.
--validateOnly run the script validation
--verbose, -v

After the command confirms the current server is the coordinator, it will attempt to find a datasource to backup. Unless --require-slave-backup has been disabled, only slaves that are ONLINE will be eligible. If no datasource can be found, the command will exit with an error. The backup will then be started on the datasource.

The cluster_backup command will wait until the cctrl command has returned before exiting. The cctrl command can return prior to the backup is completed if it takes too long or if there is another error. The tungsten_nagios_backups check or similar should be used to make sure that you always have a recent backup available in the cluster.