During failback, if prepare_failback executes successfully on the secondary companion but the primary companion fails to start, perform the following to roll back, then reissue the prepare_failback command:
Check the primary companion’s error log and the HACMP error log to find the reason the server failed to start, and correct the problems.
Stop the HACMP services on the primary node with takeover.
Log in to the secondary companion as LOGIN_NAME, and issue:
dbcc ha_admin ("", "rollback_failback")
dbcc ha_admin ("", "rollback_failover")
Both companion servers should both be back in failover mode.
Restart HACMP on the primary node.
