C Replicator (CREP) architecture is designed to tolerate the failure of different components, or connections between the components, without loss of data or loss of ability to continue replication after the failed component has been restarted.
In this topic, only loss of connection (socket calls fail or timeout) or process loss are considered failures. Performance and latency issues should be analyzed and resolved with different methods.
▪ If a target database fails, or the connection between a target and the source database fails, the application can continue writing to the source database. The transaction log keeps a record of all the events that are not replicated. The replication is left in error state. After the target database (or the connection to the source database) is restored, the replication must be manually restarted.
▪ If a source database fails, the replication is left in an error state. After the source database is restarted, the replication must be manually restarted.