Our current environment consists of a two IBM 630's in a HACMP cluster. Each node hosts multiple applications which use MQSeries. We have 'mqm' as the userid for all the Queue Managers. During HACMP failover, the MQ shutdown script runs the ipcrm command to clear out queues, shared memory and semaphores owned by the 'mqm' userid.
We want to have the ability of moving an application/queue manager to the other node if need arises, without affecting the other queue managers on the node.
We can relate the shared memory segments to the PID's of the individual queue managers, but are unable to distinguish the semaphores owned by each queue manager.
This is causing a problem because killing all semaphores effectively shuts down all queue managers on the node. Is there a way around this problem?
We've thought about creating each queue manager by a unique userid, but the semaphores/shared memory is still owned by 'mqm'.
We want to have the ability of moving an application/queue manager to the other node if need arises, without affecting the other queue managers on the node.
We can relate the shared memory segments to the PID's of the individual queue managers, but are unable to distinguish the semaphores owned by each queue manager.
This is causing a problem because killing all semaphores effectively shuts down all queue managers on the node. Is there a way around this problem?
We've thought about creating each queue manager by a unique userid, but the semaphores/shared memory is still owned by 'mqm'.