| Author |
Message
|
| velocity |
 Posted: Tue Feb 17, 2009 8:11 am Post subject: Ipprocs & CPU usage |
 |
|
Centurion
Joined: 30 Nov 2007
Posts: 126
|
Hi Guys,
The application is testing with one of the queues and i found out ipprocs is 149. This number is quite large..The server is also facing a 'cpu increase' issue which they are testing to find out the root cause. My job is to confirm it is not mq. I checked the MQ filesystems../var/mqm, /var/mqm/log..etc..everything looks fine..I honestly dont think the CPU issue is an mq issue. But the number of IPPROCS kind of scared me..Could such a high number of IPPROCS value cause memeory issue? Should the application close these handles?
Thanks in advance.. |
|
| Back to top |
|
 |
| Vitor |
| Posted: Tue Feb 17, 2009 8:19 am Post subject: Re: Ipprocs & CPU usage |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
| velocity wrote: |
| Could such a high number of IPPROCS value cause memeory issue? |
Yes, but not a serious one. Of more concern (assuming you don't really have 149 applications reading the queue) is how many connections you have to the queue manager that are dead, and hence how close you are to MaxConn (by default 200).
| velocity wrote: |
| Should the application close these handles? |
Absolutely yes. The queue manager will clean up orphaned connections eventually (a subject on which there has been much, much discussion) but it's far better for an application to close & disconnect cleanly rather than just walk away and assume "someone" will take out the garbage.
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
| mvic |
| Posted: Tue Feb 17, 2009 5:28 pm Post subject: Re: Ipprocs & CPU usage |
|
|
Jedi
Joined: 09 Mar 2004
Posts: 2080
|
| velocity wrote: |
| The server is also facing a 'cpu increase' issue which they are testing to find out the root cause. My job is to confirm it is not mq. |
It is always preferable to work towards what IS causing the problem than to start with 20 possible causes and ruling them out one by one. You have roughly 20 times as much work with the latter scheme compared with the former. Yet I see so many occasions when folks attempt to follow the latter scheme.
So the people who say the CPU is increasing - which parts of the system are taking up the CPU time? At what times of the day? What is the system supposed to be doing at those times of the day? What is that code doing? if it's doing 10 different things, then add small amounts of trace or progress reporting to the code in question (include timestamps) to find out what is happening in there.
HTH |
|
| Back to top |
|
|
| gbaddeley |
| Posted: Tue Feb 17, 2009 5:54 pm Post subject: |
|
|
Jedi Knight
Joined: 25 Mar 2003
Posts: 2538
Location: Melbourne, Australia
|
I would be concerned if there were a large number of messages on the queue and all those processes were getting messages by msg id or correl id and there was no indexing. This can cause a large cpu hit in MQ.
_________________
Glenn |
|
| Back to top |
|
|
| PeterPotkay |
| Posted: Tue Feb 17, 2009 7:19 pm Post subject: Re: Ipprocs & CPU usage |
|
|
Poobah
Joined: 15 May 2001
Posts: 7722
|
| velocity wrote: |
| The server is also facing a 'cpu increase' issue which they are testing to find out the root cause. My job is to confirm it is not mq. |
Uh, why don't the people complaining about high CPU tell you what process is taking a lot of CPU, instead everyone guessing what process might be taking a lot of CPU.
_________________
Peter Potkay
Keep Calm and MQ On |
|
| Back to top |
|
|
| velocity |
| Posted: Thu Mar 12, 2009 5:30 am Post subject: Solution |
|
|
Centurion
Joined: 30 Nov 2007
Posts: 126
|
IBM found a known problem with the QueueAgentThread (QAT) in version 6. Basically, what is happening is that the QueueAgentThread (QAT) is broswing messages, and then when it goes back to select one, it is already gone because one of the other app servers has taken it. In essence, the five app servers are competing for the same messages, which results in the 2033 messages.
The problem is fixed in MQ V7, it is also fixed in V6 on z/OS. The new QATs are smarter in that they use new options in MQ to mark the messages they have browsed, making them invisible to other QATs. The MQ JMS classes used by WAS would also need to be upgraded to V7.
However it is not feasible to upgrade at this time. Database was causing CPU increase but 2033s still exist in the traces.. |
|
| Back to top |
|
|
| kayanat |
| Posted: Tue Mar 17, 2009 9:43 pm Post subject: |
|
|
Newbie
Joined: 14 Mar 2009
Posts: 1
|
what is cpu usage exactly? whenever my cpu is running slow i notice it's really high...i visit youtube and view lots of videos often and they pause a lot at times and that's when i notice the CPU Usage is at a high percentage. ls
i checked the task manageer under processes and way at the bottom of the "Image Name" list is something that reads System Idle Process; it always has a very high number under where it says CPU...does that have anything to do with it?
________________
keyword research ~ keyword tool ~ keyword tracking ~ affiliate elite
Last edited by kayanat on Thu Mar 19, 2009 8:21 am; edited 1 time in total |
|
| Back to top |
|
|
| vol |
| Posted: Tue Mar 17, 2009 10:40 pm Post subject: |
|
|
Acolyte
Joined: 01 Feb 2009
Posts: 69
|
| i have never understood why people get upset about what they call 'high CPU usage'. as long as it does not get to 100% it is not a problem, surely? |
|
| Back to top |
|
|
| gbaddeley |
| Posted: Wed Mar 18, 2009 4:49 pm Post subject: |
|
|
Jedi Knight
Joined: 25 Mar 2003
Posts: 2538
Location: Melbourne, Australia
|
| vol wrote: |
| i have never understood why people get upset about what they call 'high CPU usage'. as long as it does not get to 100% it is not a problem, surely? |
A typical quote from an application developer? Or are you saying that in jest? Systems admin and Systems management people have a very different opinion.
100% spikes are usually OK, but 80% average is a ball park to start worrying about CPU capacity. At this point the performance some processes is probably being impacted (eg. slow response). Application and resource bottlenecks that are wasting CPU and reducing performance should be identified and relieved.
_________________
Glenn |
|
| Back to top |
|
|
| fjb_saper |
| Posted: Wed Mar 18, 2009 8:17 pm Post subject: |
|
|
Grand High Poobah
Joined: 18 Nov 2003
Posts: 20756
Location: LI,NY
|
Depends as well on the platform. Are we talking about 80% average for a mainframe or 80% average for a Unix server, for a Windows server? 3 very different things... 
_________________
MQ & Broker admin |
|
| Back to top |
|
|
| gbaddeley |
| Posted: Thu Mar 19, 2009 6:08 pm Post subject: |
|
|
Jedi Knight
Joined: 25 Mar 2003
Posts: 2538
Location: Melbourne, Australia
|
| fjb_saper wrote: |
| Depends as well on the platform. Are we talking about 80% average for a mainframe or 80% average for a Unix server, for a Windows server? 3 very different things... |
Care to enlighten us on the significant differences?
_________________
Glenn |
|
| Back to top |
|
|
| fjb_saper |
| Posted: Thu Mar 19, 2009 9:25 pm Post subject: |
|
|
Grand High Poobah
Joined: 18 Nov 2003
Posts: 20756
Location: LI,NY
|
80% average CPU in Windows and you are slowed down significantly....
80% average CPU in Unix and you are somewhat slowed down...
80% average CPU in zOS and nobody worries about it until you are hitting 100%...
_________________
MQ & Broker admin |
|
| Back to top |
|
|
| exerk |
| Posted: Fri Mar 20, 2009 12:02 am Post subject: |
|
|
Jedi Council
Joined: 02 Nov 2006
Posts: 6339
|
| fjb_saper wrote: |
| 80% average CPU in zOS and nobody worries about it until you are hitting 100%... |
Some shops might ask why are you 10% under utilised? 
_________________
It's puzzling, I don't think I've ever seen anything quite like this before...and it's hard to soar like an eagle when you're surrounded by turkeys. |
|
| Back to top |
|
|
|
|
