Author |
Message
|
Adrien |
Posted: Mon Nov 23, 2009 12:17 pm Post subject: WMB v7.0 : EG or broker freeze during deployment |
|
|
Novice
Joined: 23 Nov 2009 Posts: 14
|
Hi !
First I want to thank everyone for the great help I've always found here. I'm a WMB user since almost two years, using it to run critical business flows in my company. We are on a project to upgrade our brokers to v7.
I've been toying with this new version for a few days now and I have faced several freeze of the execution group (or even the broker) when deploying a flow via the toolkit (on windows vista). It goes very well almost all of the time but once to twice a day, the toolkit stays pending on "Deploying ...." forever. The broker is then very difficult to stop and the only mean of restoring full functionnality has been a kill/delete/recreate of the broker. Quite disturbing after the robustness experienced from the previous versions !
I'm obviously not asking for a direct solution for such a vague problem but perhaps you can provide me some thoughts ? Perhaps I need to reinstall the whole thing from a clean state.
The platform is Solaris x86-64 with MQ v7.0.1 upgraded from 7.0.0 and an Oracle 10g database as an ODBC source. The broker has been recreated after upgrading WMB from v6.1 to v7. The QM is a full repository in a cluster.
Thank you. Adrien.  |
|
Back to top |
|
 |
mqmatt |
Posted: Mon Nov 23, 2009 12:20 pm Post subject: |
|
|
 Grand Master
Joined: 04 Aug 2004 Posts: 1213 Location: Hursley, UK
|
Using the MB Explorer, take a look at the Administration Queue - are any work items listed?
Also take a look at the Administration Log in the same perspective - if something's hanging, I'd expect to see an Administration Request without a corresponding Response. |
|
Back to top |
|
 |
Adrien |
Posted: Mon Nov 23, 2009 2:08 pm Post subject: |
|
|
Novice
Joined: 23 Nov 2009 Posts: 14
|
All I can say for the moment (being away from work) is that there was some replies in the SYSTEM.BROKER.DEPLOY.REPLY which the toolkit was unable to consume.
I will try to investigate using MB Explorer tomorrow and report results back here more precisely. |
|
Back to top |
|
 |
_dave_ |
Posted: Tue Nov 24, 2009 3:49 am Post subject: |
|
|
Apprentice
Joined: 08 Apr 2005 Posts: 33 Location: IBM Hursley
|
Hi,
Did you see any errors at the broker end, in the syslog? Was the deployment successful as far as you could tell?
The extra messages on SYSTEM.BROKER.DEPLOY.REPLY could be perfectly normal. They may be for another CMP API session.
Did you try mqsistop and mqsistop -i without success? |
|
Back to top |
|
 |
mqmatt |
Posted: Tue Nov 24, 2009 3:50 am Post subject: |
|
|
 Grand Master
Joined: 04 Aug 2004 Posts: 1213 Location: Hursley, UK
|
Thanks... the messages on the S.B.DEPLOY.REPLY queue are probably unrelated to your problem; they're likely just for a toolkit session that's recently disconnected, and will be tidied up in due course. |
|
Back to top |
|
 |
Adrien |
Posted: Tue Nov 24, 2009 8:15 am Post subject: |
|
|
Novice
Joined: 23 Nov 2009 Posts: 14
|
Hi. Thank you. All your inputs are really appreciated... but i'm a bit ashamed of my question now that I discovered it was caused by a infinite loop in my flow !
The execution group went completely crazy each time we put some message in the input queue. After debugging, it finally appeared it was caused by a erroneous MRM definition which defined an element of type string with 0 length and unbounded max occurs. It was causing the MRM parser to generate a lot of elements without realizing that it was not consuming any byte in the message. After a while it eventually crashed with an out of memory error. Even if the user (me) is at fault, I was a bit surprised by this behavior and I wanted to share it with you. I was expecting a more explicit error.
But anyway, I should have checked the CPU consumption of the DataFlowEngine process before asking. Sorry. The broker is now as robust as ever. |
|
Back to top |
|
 |
mqmatt |
Posted: Tue Nov 24, 2009 8:32 am Post subject: |
|
|
 Grand Master
Joined: 04 Aug 2004 Posts: 1213 Location: Hursley, UK
|
Phew, glad you sorted it; I'll tell the dev team to stand down from panic mode  |
|
Back to top |
|
 |
|