We have an application that runs on replicated database set-up on oracle8i, with 2 master sites. One of the rep groups is synchronously replicated. Generally, when one of the DB crahes, any stored procedure call to the DB that is up from the application raises an ORA-2068 error and due to this a failover routine is invoked which removes the failed DB master site from replication. And the appln still works fine.

But when the machine itself crahes or becomes unreachable due to n/w problems, then a stored procedure call from the application to the DB that is up does not instantly throw an exception. It takes abt 15-25 mins to throw the ORA-2068 exception. The DB connection just becomes inactive for that period of time.

Can someone tell me the reason for this behaviour? Is it something to do with any sqlnet or other timeout parameters? This is critical to our appln coz it wld mean the system is actually down for 15-25 mins.

Any kind of help would be appreciated.