Rick,

I too have seen this occuring at many of my clients. What I can surmise is
that you get runaway sessions appearing with a ? when you do a 'q sess'. If
you can cancel these, then you will see that the dsmserv process CPU
utilization drops back down dramatically. I have played around a little
with idletimeout but I'm not sure if this is the correct solution.

Can someone from Tivoli provide feedback on this one? I've seen the CPU
shoot up to 100% even on SP nodes and 4-way S7A machines with lots of
memory. It will cause the machine to start thrashing. There should be a
cleaner, more automated solution instead of having to reboot the server or
manually canceling the runaway sessions.

Thanks,

Jeff Coskey
IBM Global Services
Server and Storage Solutions
3109 W. Dr. Martin L. King Jr. Blvd, Tampa, FL  33607
Phone: (813) 801-3868  T/L: 427-3868
Cell: (813) 495-6923
Pager:  (800) 759-8888 pin: 1201907
Email: coskey AT us.ibm DOT com


"Richard L. Rhodes" <rhodesr AT firstenergycorp DOT com>@VM.MARIST.EDU> on
03/08/2001 04:10:04 AM

Please respond to rhodesr AT firstenergycorp DOT com

Sent by:  "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU>


To:   ADSM-L AT VM.MARIST DOT EDU
cc:
Subject:  Runaway dsmserv


Were having a strange problem.

Over a period of several weeks we saw the cpu utilization of dsmserv rise
to
the point it was running our AIX server at 100% utilization.  It was
running
at 100% utilization even whan nothing was happening on the server - no
backups, migration, reclamation, etc.  We called support - they suggested
we
reboot our server, which was our idea also.

After the reboot, everything seemed back to normal.  Now, a week after the
reboot, dsmserv is running a constant 50% of our server, reguardless of
what's happening.  We're going to cycle dsmserv this afternoon after batch
processing.  Then, call support, again.

Is anyone else seeing this kind of behavior?

Rick