Specific properties of intra-group queuing
This section describes the specific properties of intra-group queuing including Invalidation of object handles, self recovery and retry capability of the intra-group queuing agent, and the intra-group queuing agent and serialization.
Invalidation of object handles (MQRC_OBJECT_CHANGED)
If the attributes of an object are found to have changed after the object is opened, the queue manager invalidates the object handle with MQRC_OBJECT_CHANGED on its next use.
Intra-group queuing introduces the following rules for object handle invalidation:
- If the SYSTEM.QSG.TRANSMIT.QUEUE was included in the name resolution path during open processing because intra-group queuing was ENABLED at open time, but intra-group queuing is found to be DISABLED at put time, then the queue manager invalidates the object handle and fails the put request with MQRC_OBJECT_CHANGED.
- If the SYSTEM.QSG.TRANSMIT.QUEUE was not included in the name resolution path during open processing because intra-group queuing was DISABLED at open time, but intra-group queuing is found to be ENABLED at put time, then the queue manager invalidates the object handle and fails the put request with MQRC_OBJECT_CHANGED.
- If the SYSTEM.QSG.TRANSMIT.QUEUE was included in the name resolution path during open processing because intra-group queuing was enabled at open time, but the SYSTEM.QSG.TRANSMIT.QUEUE definition is found to have changed by put time, then the queue manager invalidates the object handle and fails the put request with MQRC_OBJECT_CHANGED.
Self recovery of the intra-group queuing agent
If the IGQ agent terminates abnormally, message CSQM067E is issued and the IGQ agent starts again.
Retry capability of the intra-group queuing agent
If the IGQ agent encounters a problem accessing the SYSTEM.QSG.TRANSMIT.QUEUE (because it is not defined, for example, or is defined with incorrect attributes, or is inhibited for Gets, or for some other reason), the IGQ agent goes into the state of retry.
The IGQ agent observes short and long retry counts and intervals. The values for these counts and intervals, which cannot be changed, are as follows:
Table 1. Short and long retry counts and intervals values Constant Value Short retry count 10 Short retry interval 60 seconds = 1 min Long retry count 999,999,999 Long retry interval 1200 seconds = 20 min
The intra-group queuing agent and serialization
An attempt, by the IGQ agent to serialize access to shared queues while peer recovery is still in progress might fail.
If there is a failure of a queue manager in a queue sharing group while the IGQ agent is dealing with uncommitted messages on a shared queue or queues, the IGQ agent ends, and shared queue peer recovery takes place for the failing queue manager. Because shared queue peer recovery is an asynchronous activity, it leaves the possibility for the failing queue manager, and also the IGQ agent for that queue manager, to restart before shared queue peer recovery is complete. Which in turn leaves the possibility for any committed messages to be processed ahead of and out of sequence with the messages still being recovered. To ensure that messages are not processed out of sequence, the IGQ agent serializes access to shared queues by issuing the MQCONNX API call.
An attempt, by the IGQ agent to serialize access to shared queues while peer recovery is still in progress might fail. An error message is issued and the IGQ agent is put into retry state. When queue manager peer recovery is complete, for example at the time of the next retry, the IGQ agent can start.