| |
|
RSS Feed - WebSphere MQ Support
|
RSS Feed - Message Broker Support
|
  |
|
| Unable to get partial repository QM to refresh cluster info |
« View previous topic :: View next topic » |
| Author |
Message
|
| mattfarney |
 Posted: Tue Jan 17, 2006 7:57 pm Post subject: Unable to get partial repository QM to refresh cluster info |
 |
|
Disciple
Joined: 17 Jan 2006
Posts: 167
Location: Ohio
|
Basic structure:
Two windows Full repositories. A and B.
One windows Partial repository. C
Two Unix Partial repositories. D and E.
One of the full repos (B) is currently offline (long story, unimportant to my problem. Suffice it to say bringing it up is not currently an option.)
The cluster did not seem to be functioning properly. When I tried to
REFRESH CLUSTER, the QMs pointed at B stopped functioning.
To fix this, I tried several things: including "resolve the unknown cluster status" steps in the cluster manual. After pointing any non-repos QM pointed at B to A, I was able to resolve all of my problems except for one.
No matter what I do, C will not rejoin the cluster correctly. The repos on A sees C, but C will not find any cluster queues on A, D, or E. When I do dis clusqmgr(*), it returns itself (C) and E.
I have tried refresh cluster many times. I have removed C from the cluster and put it back in. No success.
The only other information I can add is that there are messages in the SYSTEM.CLUSTER.TRANSMIT.QUEUE on C destined for B. I have tried deleting them, but I get an error saying the queue has messages.
(Of course it has messages, I did try to delete them. =)
That message says that open processes exist (there are none by ipproc/opproc) or that there is non-snycpointed work on the QM. I even went as far as to reboot the machine, but still get this message.
My question:
How do I get C to correctly figure out it's role in the cluster?
Secondary question:
Are those messages in the SYSTEM.CLUSTER.TRANSMIT.QUEUE a problem?
Thanks!
-Matt
mfarney@csc.com |
|
| Back to top |
|
 |
| PeterPotkay |
| Posted: Tue Jan 17, 2006 8:10 pm Post subject: |
|
|
Poobah
Joined: 15 May 2001
Posts: 7722
|
issuing refresh cluster over and over rarely helps.
Having said that, on QMC, make sure the CLUSRCVR is 100% correct (conname right, clustered right), make sure the CLUSSNDR is correctly clustered and aimed at A, and issue REFRSH CLUSTER(*) REPOS(YES). The REPOS(YES) part is important in this scenario, as it will force the QM to toss any old channel defs it may know about and rejoin the cluster based on the current manual cluster channel defs.
The messages on the S.C.T.Q. that are destined for B can be cleared out if you like (delete them one by one), but more will be generated eventually, as long as B is down and know as a FR.
But do you even have a problem?
| Quote: |
The repos on A sees C, but C will not find any cluster queues on A, D, or E. When I do dis clusqmgr(*)
|
Yeah, C is a partial repository. It only knows about its own queues. Ques on other QMs wil only show up on C if C has had to get messages to them within the last 90 days. If you really did take it out of the cluster, then it lost all memory of any subscription info it may have know about.
Try doing a put on QMC to a q that is hosted on A, D or E. You may find you have no problem at all.
_________________
Peter Potkay
Keep Calm and MQ On |
|
| Back to top |
|
|
| mattfarney |
| Posted: Tue Jan 17, 2006 10:31 pm Post subject: |
|
|
Disciple
Joined: 17 Jan 2006
Posts: 167
Location: Ohio
|
I have a large set of messages in the DLQ on C. They should go to various queues on A and D (mostly). Running the DLQ generates 2085 errors.
So, it is causing an issue.
I will try a manual put tomorrow and see if it resolves.
-Matt |
|
| Back to top |
|
|
| mattfarney |
| Posted: Wed Jan 18, 2006 8:08 am Post subject: |
|
|
Disciple
Joined: 17 Jan 2006
Posts: 167
Location: Ohio
|
One thing I left off last nite:
C thinks that C and E are in its cluster (using dis clusqmgr(*))
Nothing I do can convince it otherwise. If I remove the clussdr/clusrcvr pair from C and do a refresh cluster (*) repos(YES), dis clusqmgr(*) should return an empty set.
When I did that, it continued to return C and E.
-Matt |
|
| Back to top |
|
|
|
|
|
| |
|
Page 1 of 1 |
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
|
|
|
