| Author |
Message
|
| sand |
 Posted: Sat Dec 17, 2005 6:13 am Post subject: MQ slow through application |
 |
|
Newbie
Joined: 17 Dec 2005
Posts: 4
|
We have HP-UX environment (B.11.00) with Java application server Jetty and WMQ server 5.3 (CSD 11). WMQ server is on a cluster and we are sending messages to clustered queue.
If we send two messages (data message and log message) using our custom application the whole transaction is over in one second. If the same kind of transaction comes through Jetty java application server it takes about 5 minutes. Our application has been coded with ANSI C.
The slow transaction’s WMQ trace is full of these:
14:17:13.460678 7967.1 -----{ xcsCheckProcess
14:17:13.460699 7967.1 pid(6305)
14:17:13.460715 7967.1 -----} xcsCheckProcess rc=OK
14:17:13.460724 7967.1 ----} ziiHealthCheck rc=OK
14:17:13.460734 7967.1 ----{ zcpReceiveOnPipe
14:17:13.460744 7967.1 -----{ xcsWaitEventSem
14:17:13.460761 7967.1 xcsWaitEventSem shmid:1::0::0-819416 TimeOut:10000 Type(20000)
14:17:13.460770 7967.1 ------{ xllWaitSocketEvent
14:17:13.460782 7967.1 -------{ xllSpinLockRequest
14:17:13.460790 7967.1 -------} xllSpinLockRequest rc=OK
14:17:13.460802 7967.1 -------{ xllSpinLockRelease
14:17:13.460811 7967.1 -------} xllSpinLockRelease rc=OK
14:17:13.460821 7967.1 -------{ xllListenSelectAcceptAndClose
*14:17:23.460436 7967.1 -------} xllListenSelectAcceptAndClose rc=OK
14:17:23.460467 7967.1 -------{ xllSpinLockRequest
14:17:23.460479 7967.1 -------} xllSpinLockRequest rc=OK
14:17:23.460488 7967.1 -------{ xllSpinLockRelease
14:17:23.460498 7967.1 -------} xllSpinLockRelease rc=OK
14:17:23.460507 7967.1 -------{ xllSpinLockRequest
14:17:23.460516 7967.1 -------} xllSpinLockRequest rc=OK
14:17:23.460529 7967.1 -------{ xllSpinLockRelease
14:17:23.460537 7967.1 -------} xllSpinLockRelease rc=OK
14:17:23.460546 7967.1 ------}! xllWaitSocketEvent rc=xecL_W_TIMEOUT
14:17:23.460559 7967.1 -----}! xcsWaitEventSem rc=xecL_W_TIMEOUT
14:17:23.460573 7967.1 ----}! zcpReceiveOnPipe rc=zrcC_E_TIMEOUT
14:17:23.460586 7967.1 ----{ ziiHealthCheck
Any ideas? |
|
| Back to top |
|
 |
| fjb_saper |
| Posted: Sat Dec 17, 2005 7:07 am Post subject: |
|
|
Grand High Poobah
Joined: 18 Nov 2003
Posts: 20756
Location: LI,NY
|
Can you supply the Jetty java code ?
Did you implement the JMS specific APAR to CSD11?
 |
|
| Back to top |
|
|
| sand |
| Posted: Sat Dec 17, 2005 10:50 pm Post subject: |
|
|
Newbie
Joined: 17 Dec 2005
Posts: 4
|
Thanks for the reply.
First of all both transactions goes through same WMQ’s C api (Yes, java uses c api, it’s bit complicated but needed in this case).
This slowness doesn’t generate any FDC files. I have to check the processes from the trace files for you. |
|
| Back to top |
|
|
| jefflowrey |
| Posted: Sun Dec 18, 2005 6:34 am Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
What?
You're using JNI to call MQ from Java?
That's... just wrong.
There's nothing, that I am aware of, that the C Api provides that the regular Java API does not provide...
And I would think it would be a bug if there was.
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
| wschutz |
| Posted: Sun Dec 18, 2005 6:39 am Post subject: |
|
|
Jedi Knight
Joined: 02 Jun 2005
Posts: 3316
Location: IBM (retired)
|
| Quote: |
| If the same kind of transaction comes through Jetty java application server it takes about 5 minutes. |
Hummm.... are you committing the messages right away?
_________________
-wayne |
|
| Back to top |
|
|
| sand |
| Posted: Sun Dec 18, 2005 11:13 pm Post subject: |
|
|
Newbie
Joined: 17 Dec 2005
Posts: 4
|
| We got this working. Java patches for HP-UX seemed to fix this problem. |
|
| Back to top |
|
|
| sand |
| Posted: Sun Dec 18, 2005 11:27 pm Post subject: |
|
|
Newbie
Joined: 17 Dec 2005
Posts: 4
|
| jefflowrey wrote: |
What?
You're using JNI to call MQ from Java?
That's... just wrong.
There's nothing, that I am aware of, that the C Api provides that the regular Java API does not provide...
And I would think it would be a bug if there was. |
Yes, I know it wrong.
Reason for this goes back in time 5-6 years when all needed Java components weren’t available for all needed environments (as/400 etc.) Yes, now we should have coded the whole thing with java but JNI was quicker solution. |
|
| Back to top |
|
|
|
|
