Author |
Message
|
sa2327mq |
Posted: Fri Jun 17, 2005 6:43 am Post subject: MQ Cluster repository |
|
|
Novice
Joined: 05 Jun 2005 Posts: 17
|
We run a MQ cluster for several of our apps that share queue functionality. We are planning cut-over strategy to our new hardware a question came up about repository constraints.
In a single cluster we need at least 2 repositories, but could have more.
OR
In a single cluster there can only exist 2 repositories, never more then two.
thanks |
|
Back to top |
|
 |
jefflowrey |
Posted: Fri Jun 17, 2005 6:45 am Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002 Posts: 19981
|
In a single cluster you need at least one full repository.
In a single cluster, you can have as many full repositories as you like.
In a single cluster, you rarely get any benefit from having more than two full repositories.
Full repository queue managers should be a) highly available, b) highly reachable from all queue managers in the cluster.
In some network configurations, those two reasons may provide some benefit from having more than two full repositories in a single cluster. Not much else will. _________________ I am *not* the model of the modern major general. |
|
Back to top |
|
 |
EddieA |
Posted: Fri Jun 17, 2005 7:13 am Post subject: |
|
|
 Jedi
Joined: 28 Jun 2001 Posts: 2453 Location: Los Angeles
|
And if you do have more than 2 repository managers, then each repository manager must have a manually defined CLUSSDR to every other repository.
Cheers, _________________ Eddie Atherton
IBM Certified Solution Developer - WebSphere Message Broker V6.1
IBM Certified Solution Developer - WebSphere Message Broker V7.0 |
|
Back to top |
|
 |
hguapluas |
Posted: Fri Jun 17, 2005 8:52 am Post subject: |
|
|
Centurion
Joined: 05 Aug 2004 Posts: 105 Location: San Diego
|
Shoulnd't this topic be moved to the Clustering section? |
|
Back to top |
|
 |
bbburson |
Posted: Mon Jun 20, 2005 6:41 am Post subject: |
|
|
Partisan
Joined: 06 Jan 2004 Posts: 378 Location: Nowhere near a queue manager
|
EddieA wrote: |
And if you do have more than 2 repository managers, then each repository manager must have a manually defined CLUSSDR to every other repository. |
EddieA, are you sure? I don't find this anywhere in the documentation, and my recent experience has been somewhat different.
We have recently set up a cluster in stages; two geographic locations with six queue managers in each city. I built the two locations independently with two full repositories in each site (same cluster name throughout). Then when those two clusters were fully configured and tested, I joined the two sides by defining ONE CLUSSDR in each site to point to one of the full repositories in the other site.
This seems to be working completely as I intended. Now all twelve queue managers see all the clustered queues and the messages are flowing as they should.
So it was a surprise to me to read your statement indicating that I need a few more CLUSSDR definitions. |
|
Back to top |
|
 |
sebastianhirt |
Posted: Mon Jun 20, 2005 7:22 am Post subject: |
|
|
Yatiri
Joined: 07 Jun 2004 Posts: 620 Location: Germany
|
Eddie is right. Repository information between FRs can only be exchanged via channels of the Definition Type CLUSSDRB, which means manually created ones.
Your cluster will work, but if one of the two FRs that are 'connecting' the 2 halfs of the Cluster goes bazuka, you will end up having an inconsistent Cluster.
From the Clustering Manual
Quote: |
The full repositories republish the publications they receive through the manually-defined CLUSSDR channels, which must point to other full repositories in the cluster. You must make sure that a publication received by any full repository ultimately reaches all the other full repositories. You do this by manually defining CLUSSDR channels between the full repositories. The more interconnection of full repositories you have, the more robust the cluster.
|
|
|
Back to top |
|
 |
bbburson |
Posted: Mon Jun 20, 2005 8:08 am Post subject: |
|
|
Partisan
Joined: 06 Jan 2004 Posts: 378 Location: Nowhere near a queue manager
|
sebastianhirt, thanks for the clarifications. It still sounds like interconnection of ALL full repositories is not an absolute requirement but probably should be considered best practice. |
|
Back to top |
|
 |
bbburson |
Posted: Tue Sep 12, 2006 7:36 am Post subject: Changing from auto-defined to manually-defined channels |
|
|
Partisan
Joined: 06 Jan 2004 Posts: 378 Location: Nowhere near a queue manager
|
Wow! After all this time this topic comes back up for me. I never did go back and manually define CLUSSDRs between all FRs in my cluster, and things were humming along nicely with messages flowing all around the cluster as expected.
Then one day one of the PR queue managers dropped out of the cluster for no apparent reason. Opened a PMR with IBM and the response came back "you need manually-defined CLUSSDR channels between all FRs in the cluster."
So now I have a new question: from any of my FRs I see some of the other FR queue managers with DEFTYPE(CLUSSDRA) <automatic sender definition>. If I now add manual definitions to connect these FRs, will the automatic defs go away on their own? Or will I need to stop/start channels? Or perhaps use REFRESH CLUSTER? Or (heaven forbid) bounce the queue manager?
Any guidance will be appreciated. |
|
Back to top |
|
 |
Nigelg |
Posted: Wed Sep 13, 2006 12:02 am Post subject: |
|
|
Grand Master
Joined: 02 Aug 2004 Posts: 1046
|
You get an auto CLUSSDR to all other qmgrs in addition to the manual CLUSSDR that you define. _________________ MQSeries.net helps those who help themselves.. |
|
Back to top |
|
 |
Ivans |
Posted: Wed Sep 13, 2006 2:38 am Post subject: |
|
|
Apprentice
Joined: 03 Jan 2006 Posts: 48 Location: Hursley
|
Quote: |
if I now add manual definitions to connect these FRs, will the automatic defs go away on their own? Or will I need to stop/start channels? Or perhaps use REFRESH CLUSTER? Or (heaven forbid) bounce the queue manager?
|
Defining a manual cluster sender channel (DEFTYPE=CLUSSDR), where an auto-defined cluster sender channel (DEFTYPE=CLUSSDRA) already exists, will turn the existing CLUSSDRA into a CLUSSDRB.
For more info on DEFTYPE see http://www.mqseries.net/phpBB2/viewtopic.php?t=28839&highlight=clussdrb
You do not need to start/stop channels, use REFRESH CLUSTER or bounce the queue manager
Cheers,
Ian |
|
Back to top |
|
 |
bbburson |
Posted: Wed Sep 13, 2006 5:56 am Post subject: |
|
|
Partisan
Joined: 06 Jan 2004 Posts: 378 Location: Nowhere near a queue manager
|
Thanks, Nigelg and Ivans! Just the information I needed. After I posted my question I experimented on my test cluster and came to the same conclusions. The DEFTYPE flip-flops between CLUSSDRA and CLUSSDRB and messages flow without delay. |
|
Back to top |
|
 |
|