Performance capacity for high availability

You should consider the capacity of the high availability (HA) configuration when planning for performance.

As a general rule, you should plan on an absolute maximum of 40 to 50 queue managers in an HA configuration. Be aware that if you approach this top limit, then it can take five minutes or more to restore the availability of all queue managers in the event of a failover. Too many HA queue managers also affects the performance of non-HA queue managers.

If you have more than the recommended maximum number of queue managers, as well as excessive switchover times, you might observe unwanted failed actions that require manual intervention. Note the following points:
  • The more high availability queue managers you have, the more likely you are to experience failed actions.
  • Failed actions are most likely to occur in HA switchover events, but can occur in steady state operations too.
  • With very large numbers of HA queue managers, some failed actions might require assistance from IBM Support to recover.

Expect less performance in terms of failover times and failed actions on a M2003B appliance, and as the messaging load increases.

Once configured, you should test your system in both of these conditions with your anticipated workload:
  1. A planned switchover, where you suspend one appliance in the HA group. Note that all the queue managers are stopped on the suspended appliance before they are restarted on the other appliance.
  2. An unplanned failure (for example, loss of power or loss of network). Loss of communication causes the queue managers to be started on the other appliance.