Child pages
  • Shutting Down a Cluster

This is documentation for a previous version of ClustrixDB.

Skip to end of metadata
Go to start of metadata

This information on this page applies to the Clustrix Appliance only.

 

This topic tells you how to cleanly shut down all the nodes of a cluster for storage, moving, or long periods of planned inactivity. This shutdown procedure halts the database and writes the NVRAM (non-volatile RAM) contents to a mirrored partition for redundant storage, to ensure database safety and consistency.

When a single node is powered down for more than ten minutes, the cluster begins reprotecting that node's data until fault-tolerance is restored.

To shut down a node or cluster, issue the following command:

mysql> SHUTDOWN FULL {ALL | NODE [nodeid]};

To halt a cluster of node immediately, without a clean shutdown, refer to The CLX Command-Line Administration Tool and the clx nanny commands, or simply press the power button on the nodes.

The node_id defaults to the node to which your session is connected. To shut down an entire cluster, issue the following command:

mysql> SHUTDOWN ALL;

Applies to CLX4010 ONLY

After you start this process, remove power from the system before starting up again. The NVRAM is no longer battery-backed and batteries are not re-enabled until the units are powered up again.

Very few such units exist in production. If you purchased a system from Clustrix in 2009 or later, you do not have any such units.

Log in as root and issue the following command:

node# shutdown -h now

The NVRAM is now powered down and the battery backup unit is reset so that, when you reboot, the batteries are re-enabled. Your cluster is now ready for shipment or storage.


To ensure that a cluster will not form until all the nodes are restarted and ready (avoiding partial groups):

mysql> SHUTDOWN FULL {ALL | NODE nodeid} COORDINATE_RESTART <n>;

Where <n> is the expected number of nodes in the cluster. If <n> is not provided, the expected number of nodes in the cluster will be the current quorum. 

example
 clustrix> SHUTDOWN FULL ALL COORDINATE_RESTART 5;

This statement causes '5' to be written to /etc/clustrix/expected_nodes on every machine in the cluster during the shutdown sequence. 

 

  • No labels