Re: How to identify a runaway job

[Vernon Hamberg <vhamberg@xxxxxxxxxxxxxxx>]

:) Point well-taken. I've tended to focus on prevention.

Vern

On 7/5/2013 10:42 AM, Jim Oberholtzer wrote:
> Vern:  No argument about the value of the product(s).  My contention is
> only one answer dealt with the "currently facing a major problem"  part
> of the post.  Currently means now, and a commercial product will not
> help now.  Only after installation/configuration and use will they help.
>
> Jim Oberholtzer
> Chief Technical Architect
> Agile Technology Architects
>
>
> On 7/5/2013 10:32 AM, Vernon Hamberg wrote:
> > Jim, you may be right about the critical condition. But disk/HUNTER was
> > something that'd catch this problem long before a critical condition (90
> > or 95%) appeared.
> >
> > So those solutions, as well as Martin's can often prevent the condition
> > and report on it, while getting the 95% message is about time for a
> > possible disaster.
> >
> > Vern
> >
> > On 7/5/2013 10:29 AM, Jim Oberholtzer wrote:
> > > >   Others have suggested WRKSYSACT.  You don't mention which OS level your
> > > >   running.  At V6 and above WRKSYSACT is part of the OS, at V5R4 it's part
> > > >   of the Performance tools.
> > > >
> > > >   Try just plain old WRKACTJOB and use F11 once to bring up elapsed
> > > >   numbers.  Put the cursor on the column labeled AuxIO and press F16 to
> > > >   sort.  The highest I/O job will pop to the top.  Chances are strong you
> > > >   just found your problem job.
> > > >
> > > >   Commercial products however good they are won't help when the system is
> > > >   nearly in a critical state.
> > > >
> > > >   Jim Oberholtzer
> > > >   Chief Technical Architect
> > > >   Agile Technology Architects
> > > >
> > > >
> > > >   On 7/5/2013 10:19 AM, Vernon Hamberg wrote:
> > > > > >   Denis
> > > > > >
> > > > > >   There are a couple options I can think of - one is a commercial
> > > > > >   offering, the other is free - either may help you a lot.
> > > > > >
> > > > > >   The commercial one is an app I wrote several years ago at Centerfield
> > > > > >   Technology - it may have been modified since then - disk/HUNTER was the
> > > > > >   name and is now sold by S4i and is named DASD-Plus Alert - information
> > > > > >   athttp://www.s4isystems.com/products/dasdplusalert.html
> > > > > >
> > > > > >   I forget right now who wrote the free one - they are on these lists - oh
> > > > > >   yeah, Martin Rowe - you can find info at
> > > > > >   http://www.dbg400.net/extras.html#dasdmon
> > > > > >
> > > > > >   And I recall that Help Systems has a Robot Space utility.
> > > > > >
> > > > > >   HTH
> > > > > >   Vern
> > > > > >
> > > > > >   On 7/5/2013 9:17 AM, Denis Robitaille wrote:
> > > > > > > > > >      Hello all,
> > > > > > > > > >
> > > > > > > > > >      We are currently facing a major problem. There is, on one of our systems, 1 or more job that is consuming disk at an alarming rate (the % used goes up 1 percent every 5 minutes or so). We have thousands of jobs running and we can not easily pin point the guilty one.
> > > > > > > > > >
> > > > > > > > > >      In the past, we got lucky some times (a badly defined SQL that we were able to spot), and unlucky some others (IPL when the % got to high).
> > > > > > > > > >
> > > > > > > > > >      Does anyone can give me hints on how to spot, on the fly, the disk consumption of a job, be that for Database IO or temporary memory space.
> > > > > > > > > >
> > > > > > > > > >      Thanks in advance
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > >      Denis Robitaille
> > > > > > > > > >      Chef de service TI
> > > > > > > > > >      Cascades Centre des technologies,
> > > > > > > > > >      une division de Cascades Canada ULC
> > > > > > > > > >
> > > > > > > > > >      412 Marie Victorin
> > > > > > > > > >      Kingsey falls(Québec) Canada J0A 1B0
> > > > > > > > > >      T : 819 363 6130
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > >      Pour toute information sur les événements techniques, veuillez consulter notre calendrier Groupwise CAS_CALENDRIER_TI
> > > > > >   --