MQSeries.net :: View topic - OS 390 Shared Queues acrossed two queue manager issues!

tlaner · Posted: Mon Jun 30, 2003 9:48 am Post subject:

I would appreciate any ideas about this issue we are having with Shared queues:

Let me first setup the question and later I will explain the reason we have a requirement that apparently cannot be support by MQ Series.

We have implemented shared queues across two Mainframe Queue Managers. Lets call them P3 and P7. Our configuration is primarily a Unix based middleware communicating to the Mainframe based two queue managers(P3 and P7). On the Mainframe we have two regions(mostly duplicated applications, except one application that has region affinity(can only process in one region). Also, our mainframe regions are limited to reading from one queue manager(I was told this is an IBM limitation). Because of this, the mainframe region that has the application that must process in that region only, can only read from one queue manager thus the readers are only running on the one queue manager....lets say the P3 queue manager.

Now here comes the issue. Our Unix middleware handles request messages coming from the front end. The middleware also detects whether the mainframe queue readers are up and running by inquiring for open input count(handles) on a queue. If our Unix Middleware connects to the same queue manager that the P3(region affinity) application connects to it works fine. Because the readers are local to the queue manager in question. However, we have two queue managers that Unix middleware uses. When the Unix middleware inquires for open input count for the shared queue and the queue manager is P7(meaning not the queue manager where the MF readers are running for the shared queue) it gets an open input count(mainframe readers) of 0. Thus the Unix middleware cannot determine if there is some process in the mainframe to read the queues. In this case we will not send messages because we don't want the messages to stale out.

The way our infrastructure is set up it it(will) be extremely undesirable and counter-productive to have a single point of failure(just one queue manager).

Does anyone have any suggestions on how we can get past this issue while still having two queue managers? If so, I would greatly appreciate it!

Thanks in advance,

Thomas Lane
Senior Manager, Unix Development - Middleware
Pershing Technology Group
East Brunswick, NJ