Best practices for backing up a HBase cluster
Before backing up a HBase cluster, consider the following:
Before you execute a backup job, ensure for a successful ping response from the backup hosts to hostname (FQDN) of all the nodes.
Update the firewall settings so that the backup hosts can communicate with the HBase cluster.
Ensure that the HBase table you want to protect is Snapshottable.
HBase table folder should not be deleted from hdfs if snapshot was taken on that table. If deleted, snapshot loses the reference and won't be able to restore or recover data from that snapshot.
Do not backup truncated or empty table. The backup job will fail.
Namespace name and table name must not be the same. The backup job will fail.
The tables specified for backup selection must not contain space or comma in their names.
table selection must be separated by colon. For example, namespace:tablename.
The tables specified for backup selection must not be empty.