Author |
Message
|
Shalini |
Posted: Sun Dec 12, 2004 11:11 pm Post subject: (Resolved)Two or One Full Repo QMGR for Cluster |
|
|
Master
Joined: 30 Apr 2002 Posts: 224 Location: India
|
Hi All,
MQ5.3 CSD5 (Solaris: Hub) Windows 5.3 CSD5 (Spokes)
Currently we are just in phase for moving our QMGR in hub and spoke from non MQ cluster to MQ Cluster environment.
We have only one non MQ Cluster Hub QMGR (which is under Solaris O.S Cluster)
Is it feasibility to have MQ Cluster with only one full repository in the hub (Solaris), which still has O.S Sun Cluster for Hub QMGR.
MQ Cluster docs and lots of posts here recommend us to have two for Full Repository, but as we have the O.S Sun Cluster in hub do we really need two F.R QMR.
Can you please let me know any pros & cons of having one QMGR as F.R in hub with Sun cluster?
Or any one had the similar situation where they managed to Have two F.R QMGR in the hub in spite of have O.S Cluster for their Hub QMGR?
Please comment/sugest
 |
|
Back to top |
|
 |
sebastianhirt |
Posted: Mon Dec 13, 2004 6:34 am Post subject: |
|
|
Yatiri
Joined: 07 Jun 2004 Posts: 620 Location: Germany
|
Hi, this is a quote out of some Trainingsmaterials, that I got from IBM.
Quote: |
It is recomended to have two repositories for availability reasons. A cluster can function with only one full repository; clearly if it were to fail then difficulties could arise.... |
In fact, we had problems because of having only one full repository, only couple of weeks ago...
Now we have 2 of them...
My feeling is, that it is a good idea to have at least 2.
Hardwareclusters do not prevent full repositories from beeing screwed up.
Hope this helps...
cheers
Sebastian |
|
Back to top |
|
 |
jefflowrey |
Posted: Mon Dec 13, 2004 7:10 am Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002 Posts: 19981
|
It is good practice to have two full repositories. That way, you can do maintenance on one of them and not interrupt cluster traffic.
Unless things have changed, there is no benefit from having more than two FRs. You don't get any more performance or availability. The only reason to have more than two is for network topology reasons - there are portions of your network that don't have good availability to the FRs. _________________ I am *not* the model of the modern major general. |
|
Back to top |
|
 |
basa_bbc |
Posted: Thu Feb 10, 2005 6:54 am Post subject: Two Full Repositories |
|
|
Apprentice
Joined: 20 Jan 2004 Posts: 35
|
Hello ,
Could any body help me in configuring Two full repositories or any doc. which explains this.
As i understand for 2 full repositories we have to define a Cluster sender channel between both the full repositories. But how does this handle the Load Balancing,
For Failover we have already planned to go ahead with Cluster Work Load exit.
Regards,
Bbc |
|
Back to top |
|
 |
PeterPotkay |
Posted: Thu Feb 10, 2005 7:03 am Post subject: |
|
|
 Poobah
Joined: 15 May 2001 Posts: 7722
|
Read the MQ Cluster Manual. _________________ Peter Potkay
Keep Calm and MQ On |
|
Back to top |
|
 |
sebastianhirt |
Posted: Thu Feb 10, 2005 7:14 am Post subject: |
|
|
Yatiri
Joined: 07 Jun 2004 Posts: 620 Location: Germany
|
Hi,
you might want to check out the clustering manual which you can oblige from the IBM homepage.
Yes it is right that you have to install Cluster Sender and Receiver Channels between the FRs. You cannot used the Automatic generated Cluster Sender channels (DEFTYPE(CLUSSDRA)), because they would not exchange Repository information.
But the FRs have not very much to do with work load balancing.
The standard work load balancing is working with the round roubin (Yes I think that's the name?!) mechanism. This is basicaly just to have more then one instances of a queue in a cluster. The messages are then distributed to the different instances.
The other way is to program a Cluster Workload Exit. Information for this is also available in the Clustering Documentation. This you would need to spcify then in your channels. But as I have never used this, this is pretty much all I know about it.
Hope this helps
Sebastian |
|
Back to top |
|
 |
jefflowrey |
Posted: Thu Feb 10, 2005 7:17 am Post subject: Re: Two Full Repositories |
|
|
Grand Poobah
Joined: 16 Oct 2002 Posts: 19981
|
basa_bbc wrote: |
For Failover we have already planned to go ahead with Cluster Work Load exit. |
I don't think that will really provide failover. _________________ I am *not* the model of the modern major general. |
|
Back to top |
|
 |
basa_bbc |
Posted: Thu Feb 10, 2005 7:19 am Post subject: |
|
|
Apprentice
Joined: 20 Jan 2004 Posts: 35
|
You cannot used the Automatic generated Cluster Sender channels (DEFTYPE(CLUSSDRA)).
Sebastain,
You mentioned this in the previous mail. Does this also mean that for repository information to be exchanged b/w FR and PR there has to be a defined Cluster Sender Channel and Cluster Receiver channel is it? or just defining a Cluster Receiver channel will do the trick for the repository.
Regards,
BBC |
|
Back to top |
|
 |
jefflowrey |
Posted: Thu Feb 10, 2005 7:24 am Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002 Posts: 19981
|
basa_bbc wrote: |
Does this also mean that for repository information to be exchanged b/w FR and PR there has to be a defined Cluster Sender Channel and Cluster Receiver channel is it? or just defining a Cluster Receiver channel will do the trick for the repository. |
Please read the MQ Cluster manual. _________________ I am *not* the model of the modern major general. |
|
Back to top |
|
 |
sebastianhirt |
Posted: Thu Feb 10, 2005 7:50 am Post subject: |
|
|
Yatiri
Joined: 07 Jun 2004 Posts: 620 Location: Germany
|
Quote: |
You cannot used the Automatic generated Cluster Sender channels (DEFTYPE(CLUSSDRA)), because they would not exchange Repository information |
That's what I said. You have to manualy define them. The automatic ones are not doing the trick.
Here you find further information.
http://www-306.ibm.com/software/integration/mqfamily/library/manualsa/
@ jefflowrey
I agree, it is not a failover. But I think you can cover some 'this could go wrong' scenarios with it. |
|
Back to top |
|
 |
|