MQSeries.net :: View topic

DJGoodrich · Posted: Fri Jun 14, 2002 7:52 am Post subject:

A client has established two clusters, each contains an AIX queue manager and multiple NT queue managers. The current configuration calls for HACMP to control fail-over. Under normal conditions, i.e. controlled fail-over, things appear to be working as expected.

During 'hard' failover testing, pulling disk controller cables, network connections, the HACMP config detects and makes the appropriate switch, but upon restart of the AIX queue manager on the backup server, the applications is noticing the cluster is not accepting messages. A manual 'refresh cluster' from the NT GUI appears to fix things and messages then flow as expected.

We completed the last test late yesterday, and this morning the admin had to issue the refresh in order to get the cluster back to normal.

I see the repository queues are set for non-persistent messages, but I would expect that clustering control algorithms to automatically refesh the cluster at start-up, or at least within a short period of time.

I see the options at this time of changing the persistence of the reposiroty queues, and/or forcing a 'refresh' via a script when the queue manager comes up on the backup server.

thanx in advance...