Re: Infinite loop?

[Dieter.Bender@xxxxxxxxxxx (Dieter Bender)]

On Monday 23 December 2002 04:15, you wrote:
> Well, that could be irritating.
>
> When you dump the process are you using SST (STRSST, and subsequent
> options)? Also doing wrkactjob, finding job, option 5 and then option 20
> and then option 10 on thread running shows nothing (is that what you were
> saying below with "impossible to show the thread info...")?

I have tried the WRKACTJOB approach and this brought up a time out.

>
> Is there other activity on the system - perhaps activity which runs at a
> better priority that WAS (WAS runs at priority 25 by default) at the time
> or just prior to when the CPU usage spikes? Maybe large database activity
> form the app server itself?

no, there is no other activity on the box besides WAS and me looking what the
job is doing. There is no database activity, caused by the application server.

>
> Anything in the WAS log files at the time of CPU usage spike?

No, I didn't find any information in the logs. (stdin, stdout and the trace
log)
>
> Have you noticed if the number of threads decreases to 2 (one, the
> main/first thread, waiting and a second thread running), when you do the
> endjob?

no, it doesn't. the thread with the high cpu activity (> 98%) is one of the
about 30 threads (depends on history), sometimes number 22, sometimes 16 or
another.

>
> Are you up to date on all  other PTFs (3.5.6 is of course the latest level
> of WAS)?

we have 3.5.6. installed and for the jdk we have revision 11 (august 2002),
current is 13 (december 2002)

>
> Frances Stewart
> WebSphere Application Server for iSeries 400
> IBM Rochester
>
>
>
>
>                       Dieter.Bender@t-o
>                       nline.de (Dieter         To:
> java400-l@midrange.com Bender)                  cc:
>                       Sent by:                 Subject:  Re: Infinite loop?
>                       java400-l-admin@m
>                       idrange.com
>
>
>                       12/22/2002 02:52
>                       PM
>                       Please respond to
>                       java400-l
>
>
>
>
>
> Frances,
>
> On Sunday 22 December 2002 15:29, you wrote:
> > Ito might be that the garbage collector has kicked in for some reason
>
> (such
>
> > as alot of stuff to collect) and is trying to cleanup. I have seen sone
> > case where the job appeared like it would not end, but in actuality it
>
> took
>
> > to 14 hours, because the collector was returning heap to the system (that
> > is the simplified form of what was happening).
> >
> > A process dump of all the threads (or the single thread) when the job is
> > consuming the CPU would show whether it was the garbage collector or not,
> > and also would show the stack for the thread in question.  Have done a
>
> the dump fails with timeout and it's impossible to show the thread
> information with the same problem.
>
> > WRKSYSSTS to see if storage is being returned (i.e. percent used is going
> > down) or if there is an inordinate amount of paging etc. going on in the
> > storage pool in which WAS runs?
>
> no paging, no consumption of storage.
>
> > What I would recommend ideally is that you talk to IBM Support about the
> > problem and get things setup so that when the problem occurs (hopefully
> > during normal working hours), IBM Support can get on the system and debug
> > the problem.
>
> And here starts my problem, we are in germany and ibm support tells that
> there must be a problem with the application and WAS performs fine. And
> they
> are telling we should write a little program wich makes the problem
> reproducable.
>
> Dieter
>
> > Frances Stewart
> > WebSphere Application Server for iSeries 400
> > IBM Rochester
> >
> >
> >
> >
> >                       Dieter.Bender@t-o
> >                       nline.de (Dieter         To:
> > java400-l@midrange.com Bender)                  cc:
> >                       Sent by:                 Subject:  Infinite loop?
> >                       java400-l-admin@m
> >                       idrange.com
> >
> >
> >                       12/22/2002 04:39
> >                       AM
> >                       Please respond to
> >                       java400-l
> >
> >
> >
> >
> >
> > Hi,
> >
> > we use WebSphere 3.5.6 running on as400 V5R1;
> > from time to time CPU usage reaches near 100% all used by one thread of
>
> the
>
> > application server. Nobody is working with the application and only
> > PWRDWNSYS
> > *immed brings down WebSphere. It's impossible to end the server from the
> > WepSphere console, or to stop it in another way. Neither endjob, normal
>
> or
>
> > abnormal, nor endsbs kills WebsFear.
> > It looks like an infinite loop, maybe in the JIT compile. CRTJVAPGM to
> > clearify this is a little bit complicated, beacause we use JSPs and it is
> > running hours to days.
> > The problem doesn't occur running WebSphere on another box (NT).
> >
> > any ideas?

--
mfG

Dieter Bender


DV-Beratung Dieter Bender
Wetzlarerstr. 25
35435 Wettenberg
Tel. +49 641 9805855
Fax +49 641 9805856
www.bender-dv.de