I did not see messages on the XMITQ at this time. I neglected to mention that these LPARs are on the same machine. It didn't seem like the XMITQ or channel was the issue because other messages on the "path" experienced no delay.
The client is waiting on a reply (30 second wait, three minute expiry). Even so, every message from this app took 30 seconds just to reach Sysa.
The only thing unique to this app is separate websphere address spaces and dedicated request/reply/init queues. Every other element is shared/common. When websphere restarts didn't work we "redefined" the queues. I had no expectation that it would work but it was a last resort before doing something that would affect everyone.
Paul Peplinski
| Richard Tsujimoto <[EMAIL PROTECTED]>
Sent by: MQSeries List <[email protected]> 12/07/2005 10:01 AM
|
|
Are the messages on the xmitq during the 25 sec. delay? Given that you
stated other apps take the same path, I suspect this is not the case and it
is probably due to an app issue. Is it possible that the requesting app is
expecting a response in a particular sequence and, if it doesn't get what it
expects, it delays the next request? The CICS app may have done a
destructive get, and was not able to put the response before CICS crashed.
----- Original Message -----
From: "Paul Peplinski" <[EMAIL PROTECTED]>
To: <[email protected]>
Sent: Wednesday, December 07, 2005 10:07 AM
Subject: MQ performance Issue
> List,
>
> System "A" configuration = z/OS 1.4, MQ 5.3, CICS/TS 2.3
> System "B" configuration = z/OS 1.6, MQ 5.3, Websphere 5.1
>
> Application is Sysb websphere/JMS client doing MQ request/reply with Sysa
> CICS.
>
> SYSA CICS took a TS domain abend and was recycled. Upon restarting, MQ
> messages from SYSB to SYSA (both have queue managers) for this client took
> 25 seconds to arrive (as calculated in remaining MQMD-Expiry). Nothing in
> the queues, channels, etc looked unusual. In fact other apps using the
same
> Sysa/Sysb MQ "path" were arriving instantaneously.
>
> Repeated attempts of restarting websphere control and servant address
> spaces did nothing. Out of desperation the queues used by this app were
> deleted and redefined (req q on SYSA, rep Q on SYSB (which is defined as
> remote on SYSA) and everything was fine.
>
> What could deleting and defining the queue clean up? One anomaly is that
> the "use count" on the reply queue was one at the time, even though all
> apps using it were stopped.
>
> If something was hung up, where/how do you see it? Neither MQ or Mainview
> MQ gave me anything to go on.
>
> Paul Peplinski
>
> Instructions for managing your mailing list subscription are provided in
> the Listserv General Users Guide available at http://www.lsoft.com
> Archive: http://listserv.meduniwien.ac.at/archives/mqser-l.html
>
Instructions for managing your mailing list subscription are provided in
the Listserv General Users Guide available at http://www.lsoft.com
Archive: http://listserv.meduniwien.ac.at/archives/mqser-l.html
Instructions for managing your mailing list subscription are provided in the Listserv General Users Guide available at http://www.lsoft.com
Archive: http://listserv.meduniwien.ac.at/archives/mqser-l.html
