Understanding Sun Cluster HA for WebSphere MQ Fault Monitor

Use the information in this section to understand Sun Cluster HA for WebSphere MQ fault monitor. This section describes the Sun Cluster HA for WebSphere MQ fault monitor's probing algorithm or functionality, states the conditions, messages, and recovery actions associated with unsuccessful probing, and states the conditions and messages associated with unsuccessful probing.

For conceptual information on fault monitors, see the Sun Cluster Concepts Guide.

Resource Properties

Sun Cluster HA for WebSphere MQ fault monitor uses the same resource properties as resource type SUNW.gds. Refer to the SUNW.gds(5) man page for a complete list of resource properties used.

Probing Algorithm and Functionality

• WebSphere MQ Manager

  • Sleeps for Thorough_probe_interval

  • Connects to the Queue Manager, creates a temporary dynamic queue, puts a message to the queue and then disconnects from the Queue Manager. If this fails then the probe will restart the Queue Manager.

  • If all Queue Manager processes have died, pmf will interrupt the probe to immediately restart the Queue Manager.

  • If the Queue Manager is repeatedly restarted and subsequently exhausts the Retry_count within the Retry_interval then a failover is initiated for the Resource Group onto another node.

• Other WebSphere MQ components (chi, csv & trm)

  ---

  Note –

  The probing algorithm and functionality for the Channel Initiator, Command Server and Trigger Monitor all behave the same. Therefore the following text simply refers to these components as resource.

  ---

  • Sleeps for Thorough_probe_interval

  • Dependent on the Queue Manager, if the Queue Manager fails the resource will fail and get restarted after the Queue Manager is available again.

  • If the resource has died, pmf will interrupt the probe to immediately restart the process.

  • If the resource is repeatedly restarted and subsequently exhausts the Retry_count within the Retry_interval then a failover is not initiated onto another node because Failover_enabled=FALSE has been set, the resource will instead be restarted. However, this behavior is scheduled to be enhanced at a later date to allow the resource to be marked as Offline instead of repeatedly restarted.

• WebSphere MQ Listener

  • Sleeps for Thorough_probe_interval

  • Check if the runmqlsr process associated with the Queue Manager and Port is running.

  • The listener can accommodate several port numbers under the same pmftag. If a listener for a particular port is found to be missing, the probe will initiate a restart of that listener without affecting the other listeners.

    So, although the resource can accommodate several listeners, all listeners would need to have failed before the resource is restarted. This provides a granular restart mechanism for a resource that has several listeners running.

  • If the resource is repeatedly restarted and subsequently exhausts the Retry_count within the Retry_interval then a failover is not initiated onto another node because Failover_enabled=FALSE has been set, the resource will instead be restarted. However, this behavior is scheduled to be enhanced at a later date to allow the resource to be marked as Offline instead of repeatedly restarted.