we have a customer experiencing unexpected behaviour of BSA on a RAC cluster.
Normally, BSA has been configured to run with the TAF (Transparent application failover) mechanism on a three nodes RAC cluster.
When the first node fails, the transactions are all directed on the second nodes immediately.
After having restarted the first node, if the second node fails, then it takes around ten minutes before the transactions are redirected on the first node, the active one. Very bad when a Job is running ....
Did you encounter already the same behoaviour ?
Bsa does not support TAF. If a node fails, in-flight transactions will fail, new connections will be established and subsequent transactions will hit the new node.