| Author |
Message
|
| Gerald Ball |
 Posted: Wed Feb 09, 2005 1:23 pm Post subject: Removing Cluster Definitions From Partial Repository QMGRs |
 |
|
Newbie
Joined: 09 Feb 2005
Posts: 6
|
I defined a cluster network with 2 full repositories and 5 partial repositories. Gremlins came and blanked out the REPOS() field in the full repository queue manager A and in several of the channels to it. Full repository B still regards A as a full repository queue manager, but A and the other 3 queue managers in the cluster regard it as a partial repository.
Basically queue manager A is dead in the water as a cluster queue manager. It no longer sends cluster messages - or applications messages destined for other queue managers in the cluster - and does not process cluster or applications messages that it receives. I can and have deleted the manually defined channels and queues, but the cluster queues and auto - defined channels do not go away, even with the "refresh" and "reset" commands.
Can anyone tell me how to get out of this condition? Thank you. |
|
| Back to top |
|
 |
| csmith28 |
| Posted: Wed Feb 09, 2005 3:34 pm Post subject: |
|
|
Grand Master
Joined: 15 Jul 2003
Posts: 1196
Location: Arizona
|
This post should prolly be on the Clustering Forum. Don't repost it though.
Damn those Gremlins.
I'm thinking/hoping this isn't a production environment.
I would try to suspend B then delete and rebuild it, re-insert it to the Cluster and make sure it's working. Then if A is still "dead in the water" suspend A, delete and rebuild it too.
Then refresh and reset everything.
_________________
Yes, I am an agent of Satan but my duties are largely ceremonial. |
|
| Back to top |
|
|
| PeterPotkay |
| Posted: Wed Feb 09, 2005 5:48 pm Post subject: |
|
|
Poobah
Joined: 15 May 2001
Posts: 7722
|
On A, make sure all your queues and channels are defined properly. Make sure you have a proper CLUSRCVR, and make sure you have manual CLUSSNDR to the other Full Repository.
On B, make sure it has the proper CLUSRCVR and the proper CLUSSNDR to A.
On A, make sure the Repository Name and Repoistory Namelist attributes are blank (i.e., you don't want A to be a Full yet).
On A, run (from the command line - NOT MQExplorer!!!!)
| Code: |
REFRESH CLUSTER(MyClusterName) REPOS(YES)
|
On A, make it a Full Repository by specifying the cluster name for the QM's Reposiroty attribute.
You need to run the REFRESH command with the REPOS(YES) attribute, which can only happen if A is a partial. This will give A an absolute fresh start in the cluster, at which point you can make it a Full again.
Do NOT use MQExplorer for any clustering administration. It is a royal pain in the arse when it comes to clusters. Use MO71 or the command line.
I assume you did not SUSPEND A at some point?
Do not use RESET. It is wrong here and only causing more problems.
I would not do what Chris recomended (sorry Chris).
_________________
Peter Potkay
Keep Calm and MQ On |
|
| Back to top |
|
|
| csmith28 |
| Posted: Wed Feb 09, 2005 6:22 pm Post subject: |
|
|
Grand Master
Joined: 15 Jul 2003
Posts: 1196
Location: Arizona
|
@Gerald Ball
What part of "Don't repost it though." did you not understand?
It was just a suggestion.
How about this suggestion, how about, "Don't post this on all the forums to avoid pissing everyone off." Eventually your question will be moved to the appropriate Forum. Can you tell that I've been having a bad day?
Moderators, please move this thread to the Approptriate Forum.
@PeterPotkay
Well I didn't intend for the reset to be performed until after A and B were stable but. Hey, no ambulance, no foul.
_________________
Yes, I am an agent of Satan but my duties are largely ceremonial. |
|
| Back to top |
|
|
| Gerald Ball |
| Posted: Thu Feb 10, 2005 9:28 am Post subject: Sorry For The Repost; What Might Have Caused It? |
|
|
Newbie
Joined: 09 Feb 2005
Posts: 6
|
@csmith28
I actually made my similar post on the clustering forum first. After checking my logs again and getting more information - about the suspend request and the repository queue manager definition somehow getting changed - I made the second post on the general forum, which was the one that you all saw and responded to.
Incidentally, is anyone aware of something that might have caused the repository definition to fall out? It could have been done manually, but I sincerely doubt it since it happened in the middle of the night. The only thing related that I can see is that a suspend request was issued on one of the partial repositories, and another partial repository's cluster receiver channel went down and never recovered. This was our first experience with a cluster network in a production system, and I would hope that cluster networks are more robust than that. To answer your other question, yes it was on a production environment. |
|
| Back to top |
|
|
| csmith28 |
| Posted: Thu Feb 10, 2005 10:01 am Post subject: |
|
|
Grand Master
Joined: 15 Jul 2003
Posts: 1196
Location: Arizona
|
Did you try Peter's suggestion? Any luck?....
| Quote: |
| is anyone aware of something that might have caused the repository definition to fall out? |
Not right off the top of my head no.
Have you contacted IBM Business Critical support?
| Quote: |
| I would hope that cluster networks are more robust than that. |
Normally they are. This event is rather abnormal.
_________________
Yes, I am an agent of Satan but my duties are largely ceremonial. |
|
| Back to top |
|
|
| Gerald Ball |
| Posted: Mon Feb 14, 2005 1:22 pm Post subject: Tried Peter's Suggestion, Didn't Work |
|
|
Newbie
Joined: 09 Feb 2005
Posts: 6
|
| I probably should contact IBM Support on this one ... |
|
| Back to top |
|
|
|
|
