| |
|
RSS Feed - WebSphere MQ Support
|
RSS Feed - Message Broker Support
|
  |
|
| Cluster Documentation |
« View previous topic :: View next topic » |
| Author |
Message
|
| duffMan |
 Posted: Tue Mar 23, 2004 8:04 pm Post subject: Cluster Documentation |
 |
|
Voyager
Joined: 03 Jun 2002
Posts: 75
|
Does anybody know of any documentation about MQ Clusters?
No..I'm not talking about the good 'ol IBM "smile and nod" Cluster manual. I'm talking an "insider's" guide to clustering, a "what's really going on", if you will.
DB2 has some great material available which will tell you which byte on which page on which VSAM file does what.
The one thing that p's me off about clustering is that things seem to "just happen". This morning in our production environment one of our queue managers just happened to "excuse" itself from the cluster, with abosolutely no changes (at least at the hand of a human) in the past month or so. Up until today it had been running solid for neary 2 years without a glitch. IBM tells me to remove the qmgr and add it back...a valid solution...but how did it get like that in the first place? And of course I can't recreate the problem in development because I don't understand the problem in the first place. So I just turn a blind eye and tell my director...errr...uh...um..yeah....thousands you say....errr...umm...a butterfly flapped it's wings in Toyko..
I love clustering but it's really like a bull..errr...no a kitten in a china shop.
...anyway a bit of a vent, but I'd still like to understand the internals of how a cluster maintains itself without running a battery of tests trapping repository interaction messages etc... If anyone has the inside scoupe I'd love to hear.
BTW the error was the gool old 2085 on SYSTEM.CLUSTER.COMMAND.QUEUE reported by the repository manager trying to put a partial repository's queue manager queue.
Both Qmgr's have been married in the cluster for two years now without problems. |
|
| Back to top |
|
 |
| PeterPotkay |
| Posted: Tue Mar 23, 2004 8:45 pm Post subject: |
|
|
Poobah
Joined: 15 May 2001
Posts: 7717
|
There are 3 ways to get more info about clusters:
1.) mqseries.net and the listserve
2.) The 4 day Advanced IBM Clustering Class, which gives you a very good class manual to keep
3.) The Advanced Clustering Session at the IBM MQ & CICS conferance, which gives a handout to keep.
Short of that, its just playing with it.
I know what you mean about the "magic" of clustering. Without knowing everything about what's going on under the covers, its hard to fix things, or feel confident about having to fix future problems.
I was toying around with the idea of replacing a Hub and Spoke system with a super cluster. Without a Hub, there is no longer a single point of failure. With everything in the cluster, there is no need for a gateway out of the cliuster, again a single point of failure (yes we currently have them in a MSCS hardware cluster, but it is still not 100% availability).
But I still do not feel 100% confident that I can solve most problems quickly and easily in a cluster, where as I am very confortable fixing problems with "regular" channels and transmit queues. Although in the past 6 months we have not had a single probelm with clusters, probably both a function of moving to 5.3 and practice.
| Quote: |
This morning in our production environment one of our queue managers just happened to "excuse" itself from the cluster, with abosolutely no changes (at least at the hand of a human) in the past month or so.
|
You know if someone came to you and said that about some other peice of technology, you waould be inclined to say computers don't make mistakes, people do. I gotta think your problem with the cluster after such a long period of time of stability was the result of a human mucking around where they shouldn't have. What version of MQ are you at? Clustering at 5.3 is a lot more stable than 5.2.
_________________
Peter Potkay
Keep Calm and MQ On |
|
| Back to top |
|
|
| JasonE |
| Posted: Wed Mar 24, 2004 1:33 am Post subject: |
|
|
Grand Master
Joined: 03 Nov 2003
Posts: 1220
Location: Hursley
|
Dont forget that clustering has a great 'feature' that something might go wrong today, but it might not show up for 28 or 90 days.. Try debugging that
I agreee with Peter though, from a service point of view 5.3 clustering is looking really good (eg. at fp5). Even the later CSDs on 5.2 seemed to improve things quite nicely.
If you want to look at the internals, amqrfdm is a great place to start as it lists out what the running qmgr thinks of the system at the moment. However, its completely undocumented so dont expect any help |
|
| 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
|
|
|
|
