FAILOVER_TX_TIMEOUT configuration parameter

In high-availability cluster environments, use the FAILOVER_TX_TIMEOUT configuration parameter to enable transactions to complete after failover of the primary server.

Use the FAILOVER_TX_TIMEOUT configuration parameter to indicate the maximum number of seconds after failover that the server waits before it begins rolling back transactions. Set the FAILOVER_TX_TIMEOUT configuration parameter to the same value on all servers in a high-availability cluster.

onconfig.std value
FAILOVER_TX_TIMEOUT 0
takes effect
After you edit your onconfig file and restart the database server.
When you reset the value dynamically in your onconfig file by running the onmode -wf command.
When you reset the value in memory by running the onmode -wm command.

Usage

When a failover occurs in a high-availability cluster environment, one of the secondary servers takes over the role of the primary server. The secondary server that becomes the new primary server is called the failover server.

You enable transaction survival by setting the FAILOVER_TX_TIMEOUT configuration parameter to a value greater than zero. When transaction survival is enabled, the failover server must be able to contact the remaining secondary servers to synchronize and resume any open transactions. Similarly, the surviving secondary servers must be able to establish connections to the failover server to re-send any pending transactions. The FAILOVER_TX_TIMEOUT configuration parameter specifies how long the servers wait before they begin rolling back transactions.

On the failover server, if the number of seconds specified by FAILOVER_TX_TIMEOUT is exceeded, any open transactions that are not synchronized with a surviving server are terminated and rolled back.

On the remaining secondary servers, if the number of seconds specified by FAILOVER_TX_TIMEOUT is exceeded, any open transactions on that server return an error.

Set FAILOVER_TX_TIMEOUT to 0 to immediately roll back all open transactions when failover occurs.

If the primary server fails and a secondary server fails to take over the role of the primary server, then any open transactions are rolled back, and the client is unable to make updates. For example, if an update activity has been started on a secondary server and the primary server fails, and then that failover processing does not complete and a new primary server is not established, after a predetermined amount of time, the client request times out, placing the sqlexec thread in an indeterminate state.

In the preceding scenario, active transactions are rolled back, but the physical rollback cannot occur until the new primary server is established (because the primary server manages the logs). Under these circumstances, the session can be unaware of operations that were performed on the secondary server. The session can be unaware of the rollback of a partially applied transaction because the rollback of the partial transaction cannot occur until a new primary server is established.


Copyright© 2019 HCL Technologies Limited