If a primary database fails and you are unable to recover all committed transactions, you must load the database to a previous state and follow a recovery procedure designed to restore consistency at the replicate sites.
Most database failures are recovered without losing any committed transactions. No special Replication Server recovery procedure is needed if the database recovers on restart—Replication Server performs a handshake with the database, ensuring that no transactions are lost or duplicated in the replication system.
Recovering with primary dumps only.
If you do not have coordinated dumps, you can load the failed primary database and then verify the consistency of the replicate databases with the restored primary database.
Recovering with coordinated dumps.
If you have coordinated dumps of primary and replicate databases, you can use them to load all databases in the replication system to a consistent state.