| |
|
RSS Feed - WebSphere MQ Support
|
RSS Feed - Message Broker Support
|
  |
|
| JCA client starvation? |
« View previous topic :: View next topic » |
| Author |
Message
|
| jaxzin |
 Posted: Sun Nov 07, 2010 9:13 pm Post subject: JCA client starvation? |
 |
|
Newbie
Joined: 07 Nov 2010
Posts: 3
|
I'm using Websphere MQ 7.0.1.0 and connecting JavaEE Message Driven
Beans to it from Glassfish 2.1 using IBM's JCA adapter. Â I have a
cluster of 4 Glassfish servers running an MDB (400 pooled instance total)
that all read off one queue. Â Oddly, the CPU and I/O charts for the 4 boxes
are not uniform and one appears to do more work than the others, but
its not consistent which box does more work, it changes week-to-week.
Is what I'm doing inherently wrong? How do I diagnosis the issue? Â Any
help would be great.
Thanks,
Brian |
|
| Back to top |
|
 |
| fjb_saper |
| Posted: Sun Nov 07, 2010 10:12 pm Post subject: Re: JCA client starvation? |
|
|
Grand High Poobah
Joined: 18 Nov 2003
Posts: 20756
Location: LI,NY
|
| jaxzin wrote: |
I'm using Websphere MQ 7.0.1.0 and connecting JavaEE Message Driven
Beans to it from Glassfish 2.1 using IBM's JCA adapter. I have a
cluster of 4 Glassfish servers running an MDB (400 pooled instance total)
that all read off one queue. Oddly, the CPU and I/O charts for the 4 boxes
are not uniform and one appears to do more work than the others, but
its not consistent which box does more work, it changes week-to-week.
Is what I'm doing inherently wrong? How do I diagnosis the issue? Any
help would be great.
Thanks,
Brian |
What is your value for SHARECNV on the SVRCONN channel?
I suppose that you set your MAXCHANNELS / MAXACTIVECHANNELS to at least 1,000 ... 
_________________
MQ & Broker admin |
|
| Back to top |
|
|
| PeterPotkay |
| Posted: Mon Nov 08, 2010 4:44 am Post subject: |
|
|
Poobah
Joined: 15 May 2001
Posts: 7722
|
When 2 or more MQ Clients read from a single queue, racing for the messages that become available, there is no guarantee that all the clients will have an even distribution of work.
When the clients are keeping up with the arrival rate, meaning there are 1 or more clients waiting with nothing to do, you'll frequently find that the distribution of work is far from equal. The clients don't communicate amongst themselves to see who is next in line for the next message. All you can bank on is that one of those clients will grab a message when the message is ready and you won't find yourself waiting.
If the messages are arriving faster than they are being consumed, and the queue is backing up, you will then see a relatively even work load distribution, assuming all the clients are otherwise equally capable.
_________________
Peter Potkay
Keep Calm and MQ On |
|
| Back to top |
|
|
| Vitor |
| Posted: Mon Nov 08, 2010 5:26 am Post subject: Re: JCA client starvation? |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
| jaxzin wrote: |
| Any help would be great. |
Point one: PeterPotkay has pretty much nailed the situation & I agree with his summation.
Point two: if you want the 4 boxes to have more even distribution of workload give each one it's own queue manager & instance of the queue to read & put them in a cluster.
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
| mqjeff |
| Posted: Mon Nov 08, 2010 5:28 am Post subject: Re: JCA client starvation? |
|
|
Grand Master
Joined: 25 Jun 2008
Posts: 17447
|
| Vitor wrote: |
| jaxzin wrote: |
| Any help would be great. |
Point one: PeterPotkay has pretty much nailed the situation & I agree with his summation. |
The simple solution to take away is that 400 instances is *too many*, if you see this kind of unbalanced workload at peak load. |
|
| 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
|
|
|
|
