Cluster configuration lost when machine restarts/ qmanager restarts

[ghivtt]

Hi

I am using MQ 5.3 on Windows 2K.
Using MQ Explorer, I create a cluster of queue managers across many computers. When I restart the machine where there is a full repository manager, the cluster configuration seems to be lost.

Whenever I restart the machine, I need to do the following to restore the cluster:
On the queue manager, right click and execute
1. Leave Cluster
2. Then, Join Cluster.
3. Resume Cluster (for a few times).
4. Refresh Cluster (for a few times).

If this does not work, I have to delete the queue managers and again do the entire process right from re-creating the queue managers and the cluster queues.

Why does this happen ? Is there not any way to save the cluster configuration ?

Also such type of problems occur when the repository queue manager is restarted.

Can anyone please help? Thank you.

 

[kevinf2349]

Do you only have one repository qmgr?

 

[ghivtt]

No, I am having 2 full repository queue managers. And one partial repositry queue manager.
Also, in the MQ Explorer, in queuemanager properties, many times, the Repository tab shows Repository for Cluster but the cluster tab does not show entry "Yes" against the Repoistory column.

 

[mqonnet]

Most likely your listeners are not started by default when you start your qm and hence you dont see any cluster sdr/rcvr channels running and which would lead you to believe that clustering is not enabled anymore on this qm.

Check to see if your cluster sdr/rcvr definitions are still intact and that you have your listener running.



Cheers
KK

 

[ghivtt]

Thanks mqonnet.
You are right. The cluster channels are not started when the machine restarts.
I am trying to use the channel service to start it automatically.
Also the listener needs to be running.

Doubt1:
I found that although these services are automatic some times they do not start automatically. What might be the reason?

Doubt2:
If I change the short/long retry interval of the sender channel, still it uses the default one (60 sec for short and 12 mins for long). I restarted the qmgr, channle etc. still it does not use the new ones(I put the new values as 10 secs for both intervals for quick detection of failure).

Any help on these queries will be highly appreciated.

Thanks in advance to all of you (and the hosts od tek tips.com)

 

[mqonnet]

Doubt 1. You have to change i think in the services or the registry to make it automatic.

Doubt 2. Any changes to the cluster take effect only when you "REFRESH THE CLUSTER" using refresh cluster command. This should get you the new values for the channels that you defined.



Cheers
KK

 

[pmeekin]

Check if you have messages stuck in SYSTEM.CLUSTER.TRANSMIT.QUEUE on each system. Also, check the Queue Depth of SYSTEM.CLUSTER.REPOSITORY.QUEUE before and after restarting the Repository queue manager.

The only time I've had problems like this was following some 'messing around' that I did. Delete the complete cluster and redefine it from scratch to be sure. Ensure that all QMgrs can connect using the CONNAME parameters given on all cluster channels.

Cheers,
Paul