| Author |
Message
|
| sebastia |
 Posted: Sat Sep 29, 2007 8:07 am Post subject: cluter sender channel uses (old) wrong port number |
 |
|
Grand Master
Joined: 07 Oct 2004
Posts: 1003
|
Hi !
I have (manually) defined a CLUSTER-SENDER channel as
: DEFINE CHANNEL ('TO.qmc2') CHLTYPE(CLUSSDR) +
: CONNAME('localhost(1492)') +"
But ( in AMQERR01.LOG fom "qmc1" qmgr ) I still keep on finding
>>> The attempt to allocate a conversation using TCP/IP to
>>> host 'myhostname (╠UÑqΣUÑq)(1414)' was not successful.
Why is the channel trying to use port 1414 for the connection ?
I guess some old definition in the Cluster Repository ...
HOW DO I FIX THAT ????
Thanks. Sebastian.
======================================== |
|
| Back to top |
|
 |
| PeterPotkay |
| Posted: Sat Sep 29, 2007 8:21 am Post subject: |
|
|
Poobah
Joined: 15 May 2001
Posts: 7722
|
What is the definition of the CLUSRCVR channel on QM2? A manually defined clussndr is only ever used for the first communication to the FR, or if the refresh command is issued with the repos yes option. Otherwise auto clussndrs are defined on the fly based on the destination QM's clusrcvr definition. QM2 is a Full Repository, right?
_________________
Peter Potkay
Keep Calm and MQ On |
|
| Back to top |
|
|
| sebastia |
| Posted: Sat Sep 29, 2007 10:20 am Post subject: |
|
|
Grand Master
Joined: 07 Oct 2004
Posts: 1003
|
Yes, QM1 is Primary Repository and QM2 is PR also.
In fact, there are no more qmgrs in the cluster yet ...
( )
Yes, I did a first (and fast, so with few errors) definition
and then corrected it to have the specific port numbers
but this new values are not "propagated" to the cluster.
They are visible if you look at the local Queue Manager (in MQ Explorer)
but the value in the "Cluster view" in MQ Explorer does not change.
REFRESH command ? I shall try ...
"Now that I know the answers, they change the questions ..." |
|
| Back to top |
|
|
| PeterPotkay |
| Posted: Sat Sep 29, 2007 11:30 am Post subject: |
|
|
Poobah
Joined: 15 May 2001
Posts: 7722
|
There is no such thing as a "Primary" repository. Use the terms Full or Partial. If you shorten Primary to PR like you did most people will read that as Partial and I don't think that's what you intend.
If your cluster channel defs are 100% correct now, alter QM1 to be a partial repository and issue REFRESH CLUSTER with the REPOS(YES) option. Then make it a Full Repository again. Do the same on QM2.
(You can't use the repos yes option on a full repostotory so that's why you temporarily make it a partial when you have to run this command on a Full)
_________________
Peter Potkay
Keep Calm and MQ On |
|
| Back to top |
|
|
| fjb_saper |
| Posted: Sat Sep 29, 2007 11:41 am Post subject: |
|
|
Grand High Poobah
Joined: 18 Nov 2003
Posts: 20756
Location: LI,NY
|
| PeterPotkay wrote: |
There is no such thing as a "Primary" repository. Use the terms Full or Partial. If you shorten Primary to PR like you did most people will read that as Partial and I don't think that's what you intend.
If your cluster channel defs are 100% correct now, alter QM1 to be a partial repository and issue REFRESH CLUSTER with the REPOS(YES) option. Then make it a Full Repository again. Do the same on QM2.
(You can't use the repos yes option on a full repostotory so that's why you temporarily make it a partial when you have to run this command on a Full) |
Before that make sure you stop and restart any of the channels that show up different in the status from their definition.
Enjoy 
_________________
MQ & Broker admin |
|
| Back to top |
|
|
| sebastia |
| Posted: Sun Sep 30, 2007 1:14 am Post subject: |
|
|
Grand Master
Joined: 07 Oct 2004
Posts: 1003
|
Peter - you are right (again) - bith are Full Repository.
The actual (test envir) design is 2 qmgrs, both of them Full Rep. The final design will be 6 qmgrs, and we plan to use only 2 Full rep.
Today I have learnt : I shall NOT "shorten" any names, so no trouble will be induced.
"REFRESH CLUSTER with the REPOS(YES) option" - on a "Partial Repos" : I shall read carefully that part of the manual.
Saper : one channel "TO.qm1" is "Running" on both sides. The other, "TO.qm2" is "Retrying" on sender, and "Inactive" on receiver side.
I guess it's better "refresh" the cluster with all in "stopped" state.
Have a nice weekend. Sebastian. |
|
| Back to top |
|
|
| fjb_saper |
| Posted: Sun Sep 30, 2007 1:33 am Post subject: |
|
|
Grand High Poobah
Joined: 18 Nov 2003
Posts: 20756
Location: LI,NY
|
| sebastia wrote: |
Saper : one channel "TO.qm1" is "Running" on both sides. The other, "TO.qm2" is "Retrying" on sender, and "Inactive" on receiver side.
I guess it's better "refresh" the cluster with all in "stopped" state.
Have a nice weekend. Sebastian. |
No. When you refresh the cluster the cluster sender and cluster receiver should be running, at least between the PR and its FR.
When having a clustersdr channel in retry mode, check first the definition of the corresponding clusterrcvr on the target. Remember the definition of the clusterrcvr must be valid throughout all members of the cluster...
Always stop a channel before running the alter chl command. Restart it afterwards. You will need to stop and restart the cluster sender chl on the FR for it to pick up the change...and remember it is autogenerated... so you need to either use mqsc or do it going through the channel status screens as there is no definition for that chl ( definition is the clusterrcvr from the PR)
If you set things right in the FR attached to the PR and need to briefly make the other FR a PR run refresh cluster(mycluster) repos(yes) and make it a Full Repository again, the new definition should get slowly replicated...
Any PR lagging behind, use refresh cluster repos(yes) on it.
Enjoy
_________________
MQ & Broker admin |
|
| Back to top |
|
|
| sebastia |
| Posted: Sun Sep 30, 2007 8:42 am Post subject: |
|
|
Grand Master
Joined: 07 Oct 2004
Posts: 1003
|
OK - I understand what you say : the channels must be running, not "stopped". But if the definition is not "consistent", this is, one of the channels is not "Running" but "Retrying", then there will be no communication until we "fix" the problem. So, if the "port" is wrong, I have to stop the "Sender" part of the channel, make its q mgr a Partial Repository, run the "Refresh" and then make it Full Repository again. OK - I can try this ... on Monday.
All your opinions are welcome ! Sebastian. |
|
| Back to top |
|
|
| fjb_saper |
| Posted: Sun Sep 30, 2007 3:36 pm Post subject: |
|
|
Grand High Poobah
Joined: 18 Nov 2003
Posts: 20756
Location: LI,NY
|
| sebastia wrote: |
OK - I understand what you say : the channels must be running, not "stopped". But if the definition is not "consistent", this is, one of the channels is not "Running" but "Retrying", then there will be no communication until we "fix" the problem. So, if the "port" is wrong, I have to stop the "Sender" part of the channel, make its q mgr a Partial Repository, run the "Refresh" and then make it Full Repository again. OK - I can try this ... on Monday.
All your opinions are welcome ! Sebastian. |
The problems with your channels is that you think p2p and not cluster.
What makes you assume that the if the port is wrong it's the cluster sender's fault? As you only have 2 FR I would first check the cluster senders but ALSO check the corresponding cluster receiver. In my experience a wrong port number is more often caused by a typo on a cluster receiver than on a cluster sender.
So the right sequence is
- Find the channel with the wrong definition
A cluster sender will be in retry mode.
The cluster sender might have been defined on the box. Check it has the right definition.
A cluster sender can also be autodefined on the box. In that case the cluster receiver was manually defined, probably with the wrong information. You always need to fix the channel where it was manually defined.
So if the cluster sender to.qm2 on qm1 is in retry the channel you might have to fix is the cluster receiver to.qm2 on qm2...
- stop the channel
- fix the definition
- start the channel
- check for any instances of the channel still in retry mode (across the cluster)
- recycle said instances of the channel (stop/start)
- Check everything is working ok (mainly between PR and associated FR)
- Check information known at FR level (channel, conname etc...)
- Check the information on the other FRs and apply next points to resolve
- Check for any other box having the old info (channel in retry)
- If it is an FR and you can't make it a PR and run refresh cluster repos(yes) then make it an FR again, just run refresh cluster, give it a minute or 2 and check the information again. It should have updated
- If it is a PR run refresh cluster(mycluster) repos(yes)
- reverify the information has been updated and is now current on the qmgr you are working on.
Enjoy
_________________
MQ & Broker admin |
|
| Back to top |
|
|
| jefflowrey |
| Posted: Sun Sep 30, 2007 4:49 pm Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
When working with clusters, I've never seen a lot of value in dealing with any CLUSSDRs that were not pointed at an FR. And then only to make sure that they were using the right conname information.
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
| sebastia |
| Posted: Mon Oct 01, 2007 10:21 am Post subject: |
|
|
Grand Master
Joined: 07 Oct 2004
Posts: 1003
|
I think now is the moment to ask it :
HOW DO I DO to change a FR queue manager to be PR ?
I have been looking thru (almost) all MQ Explorer and couldn't find it !
( and I know it will be stamped in the front of my face, but man,
I have been searching it, true )
Sebastian. |
|
| Back to top |
|
|
| fjb_saper |
| Posted: Mon Oct 01, 2007 2:14 pm Post subject: |
|
|
Grand High Poobah
Joined: 18 Nov 2003
Posts: 20756
Location: LI,NY
|
| sebastia wrote: |
I think now is the moment to ask it :
HOW DO I DO to change a FR queue manager to be PR ?
I have been looking thru (almost) all MQ Explorer and couldn't find it !
( and I know it will be stamped in the front of my face, but man,
I have been searching it, true )
Sebastian. |
| Code: |
alter qmgr repos(' ')
alter qmgr repos(mycluster) |
Replace repos with reposnl if you deal with multiple clusters on the same qmgr.
_________________
MQ & Broker admin |
|
| Back to top |
|
|
|
|
