9.8.3. Restoring the Cluster

Note

The restore process restores an entire cluster to a previous state.
This is not intended to restore a single node. Because all data are replicate between all nodes, it is simpler and safer to remove a failed node and install a new one than it is to attempt to restore the node.

Important

Every step must be performed on every node in the cluster.
  1. Shut down every node in the storage cluster. Run the stop command on every storage machine:
    [root@server ~]# serverRoot/jon-server-3.2.GA/bin/rhqctl.sh stop --storage
  2. Remove the commit_log/ directory for each node.
    [root@server ~]# rm * /opt/jon/rhq-data/data/commit_log/*
  3. Delete all files in the metrics_index directory, except for the snapshot files.
    [root@server ~]# rm /opt/jon/rhq-data/data/data/rhq/metrics_index/*.*
  4. Copy all files from the desired snapshot directory into the metrics_index directory.
    [root@server ~]# cp /opt/jon/rhq-data/data/data/rhq/metrics_index/snapshots/timestamp/* /opt/jon/rhq-data/data/data/rhq/metrics_index
  5. Restart each storage node.
    [root@server ~]# serverRoot/jon-server-3.2.GA/bin/rhqctl start --storage