Thanks all for the input, You are correct Rob. In my distant past we did have 4 qmgrs, 3 UNIX, and one being a OS/390. The app resided on the OS/390 sending to the clustered queues and itself (OS/390) not hosting any local queues.
I just proved out this scenario by eliminating the physical queue (local) on one of my queue managers in the cluster and left the other 2. The app is load balancing correctly now, although not a 1 to 1 ratio, which I'll need to look at. I knew it was something simple. Thanks again, Donald S. Murray MQSeries Systems Engineer Desk- 201-369-8624 -----Original Message----- From: Wyatt, T. Rob [mailto:[EMAIL PROTECTED] Sent: Wednesday, November 19, 2003 1:47 PM To: [EMAIL PROTECTED] Cc: Murray, Donald Subject: RE: Cluster workload balancing (Was: How do you test a MQSeries r ecovery plan) Don, Changed the subject of this thread for clarity. If you have three QMgrs all with local instances of the same cluster queue and your client connects to one of them, the messages will go only to the local instance of that - as specified by IBM. If you were to have a fourth QMgr in the cluster and it had no local instance of the cluster queue, putting the messages to that 4th QMgr would result in the expected round-robin behavior. Per the manual: "The workload management algorithm selects the local queue manager as the destination whenever possible." See Chapter 5 - Workload Balancing of the cluster manual at: http://publibfp.boulder.ibm.com/epubs/html/csqzah06/csqzah06tfrm.htm -- T.Rob -----Original Message----- From: Don Murray [mailto:[EMAIL PROTECTED] Sent: Wednesday, November 19, 2003 9:36 AM To: [EMAIL PROTECTED] Subject: Re: How do you test a MQSeries recovery plan Good morning or afternoon as whatever the case may be. I'm running in to a problem with MQ cluster load balancing with-out an exit. We have 10 applications on various servers connecting to a cluster (XXX_CL1). All applications are connected calling the MQ Client to one of the 3 servers and puts to a queue that is clustered on all 3 servers. 3 different qmgrs, 3 different servers, created 3 identical queues (clustered with-in the same cluster), on 3 full repositories; Default Bind is Not Fixed. Cluster senders and receivers are running between all qmgrs. All queues and clustered queues are seen through all the qmgrs. When putting messages via a client channel the messages go to one queue only and not in a round robin as specked? There is no qmgr specified in the put as the messages should resolve to the next qmgr in the cluster work load list. With that in mind, would this issue be caused by the set MQSERVER environment variable that is being used? If so how do I get around this? What am I missing? Donald S. Murray MQSeries Engineer Desk- 201-369-8624 Instructions for managing your mailing list subscription are provided in the Listserv General Users Guide available at http://www.lsoft.com Archive: http://vm.akh-wien.ac.at/MQSeries.archive Instructions for managing your mailing list subscription are provided in the Listserv General Users Guide available at http://www.lsoft.com Archive: http://vm.akh-wien.ac.at/MQSeries.archive