| Author |
Message
|
| sebastianhirt |
 Posted: Fri Aug 11, 2006 6:07 am Post subject: MB6 Deployments are taking ages |
 |
|
Yatiri
Joined: 07 Jun 2004
Posts: 620
Location: Germany
|
Hey all,
One of my Customers is having a kinda weird problem.
They are trying to deploy a flow that they have migrated from V2.1 to a MB 6.0.0.1 on Solaris.
This is taking up to 7 hours for them, which is not so cool.
The BAR file is about 18 MB (3 Message flows plus 3 message sets).
So OK. I understand that this would take a while. But that long?
I have seen problems with deployment speed if some of the new nodes are used in a message flow. But these message flows are only containing 'old' nodes.
Now the really weird part is, if they split it up in 6 different packages the 6 deployments are done in almost no time.
ConfigMgr and Broker are running on the same machine, so the MQConnection is not the bottleneck.
So I was thinking maybe it's just the box that is not big enough. But performance meassurements show that not even 50% of CPU was ever used, and that there is plenty of free memory. Apperantly there are no user limits set. So I think I can sort of exclude all the hardware.
any ideas or thoughts?
cheers
Sebastian |
|
| Back to top |
|
 |
| mqmatt |
| Posted: Fri Aug 11, 2006 7:13 am Post subject: |
|
|
Grand Master
Joined: 04 Aug 2004
Posts: 1213
Location: Hursley, UK
|
The performance of the deploy is really down to the complexity of the message flows and message sets within it. Are you saying that performance was better in v2.1?
Remember that the deployment model is completely different in v6. In v2.1 when you issued a deploy, only changed objects were actually sent down to the broker. But v6 gives you the choice... If you deploy a BAR file that contains your entire set of message flows and message sets, then everything will be deployed and it will take some time. Of course, if you're now running the CM on a Solaris machine, then it will have a different capacity (and throughput rate) compared with the v2.1 CM machine.
Like-for-like deployment should be a lot quicker than v2.1; the CM is no longer a bottleneck for message flow compilation and the interaction between the various components has been improved (fewer, smaller messages).
Cheers
-Matt |
|
| Back to top |
|
|
| shrek |
| Posted: Sun Aug 13, 2006 1:07 pm Post subject: |
|
|
Acolyte
Joined: 19 Feb 2005
Posts: 61
Location: Gudivada,India
|
Have you tried performing below
1. Remove deployed children from Execution Group
2. Perform the deploy
Just a suggestion..
Thanks. |
|
| Back to top |
|
|
| nvogt |
| Posted: Sun Aug 13, 2006 5:14 pm Post subject: |
|
|
Novice
Joined: 27 Apr 2006
Posts: 11
Location: Aachen, Germany
|
Hello Sebastian
We have had the same problem on AIX.
One of the problems is that even if the deploy has completed the mqsideploy command hangs until the configured timeout is over up to 3 times.
we have a fix from IBM for AIX maybe there will be one for solaris too.
best regards
Norbert |
|
| Back to top |
|
|
| sebastianhirt |
| Posted: Sun Aug 13, 2006 11:56 pm Post subject: |
|
|
Yatiri
Joined: 07 Jun 2004
Posts: 620
Location: Germany
|
| mqmatt wrote: |
The performance of the deploy is really down to the complexity of the message flows and message sets within it. Are you saying that performance was better in v2.1?
|
Yes Matt, that's what it comes down to. OK. The complexity of those flows is really high. But never the less, 7 hours for 3 flows and 3 sets?
| Quote: |
Like-for-like deployment should be a lot quicker than v2.1; the CM is no longer a bottleneck for message flow compilation and the interaction between the various components has been improved (fewer, smaller messages).
|
In my case it really seems to be the config manager that is the bottle neck. The timeouts for configuration in the broker are set to 20 minutes, and we don't get timeouts.
shrek,
It is even the original deploy that is taking that long-- not only on updates.
Norbert,
would you mind letting me know the PMR number that was involving your problem? I would want to interview the IBM support on this. Maybe this gives them an idea on how to fix ours.
Thanks
Sebastian |
|
| Back to top |
|
|
| nvogt |
| Posted: Wed Aug 16, 2006 1:08 am Post subject: |
|
|
Novice
Joined: 27 Apr 2006
Posts: 11
Location: Aachen, Germany
|
Hello Sebastian
Please excuse the late reply
our PMR is 34074,075,724
best regards
Norbert |
|
| Back to top |
|
|
| sebastianhirt |
| Posted: Wed Aug 16, 2006 1:19 am Post subject: |
|
|
Yatiri
Joined: 07 Jun 2004
Posts: 620
Location: Germany
|
|
| Back to top |
|
|
| mqmatt |
| Posted: Thu Aug 24, 2006 7:07 am Post subject: |
|
|
Grand Master
Joined: 04 Aug 2004
Posts: 1213
Location: Hursley, UK
|
| A cause of slow deployment problems in the Config Manager has been identified in the v6 code. The fix will (all being well) make into the next fixpack, but contact IBM if you'd like an interim fix. |
|
| Back to top |
|
|
|
|
