Re: 20001214: LDM: out of per user processes

Unidata Support wrote:
> 
> ------- Forwarded Message
> 
> >From: Tom McDermott <address@hidden>
> >Subject: LDM: out of per user processes
> >Organization: SUNY Brockport
> >Keywords: 200012141622.eBEGM4o06206 LDM processes
> 
> Hi,
> 
> This morning when I came in (delayed several hours beacause of a
> snowstorm), no users were able to access our server.  The reason for this
> was clear from these messages in the system log:
> 
> Dec 14 06:31:03 vortex unix: NOTICE: out of per-user processes for uid 214
> Dec 14 06:32:25 vortex last message repeated 23 times
> Dec 14 06:32:26 vortex unix: WARNING: out of processes
> Dec 14 06:32:28 vortex last message repeated 2 times
> Dec 14 06:32:28 vortex unix: NOTICE: out of per-user processes for uid 214
> Dec 14 06:32:29 vortex unix: WARNING: out of processes
> Dec 14 06:32:30 vortex last message repeated 2 times
> Dec 14 06:32:30 vortex unix: NOTICE: out of per-user processes for uid 214
> Dec 14 06:32:32 vortex unix: WARNING: out of processes
> Dec 14 06:32:32 vortex last message repeated 1 time
> Dec 14 06:32:33 vortex unix: NOTICE: out of per-user processes for uid 214
> Dec 14 06:32:34 vortex unix: WARNING: out of processes
> Dec 14 06:32:35 vortex last message repeated 3 times
> Dec 14 06:32:35 vortex unix: NOTICE: out of per-user processes for uid 214
> Dec 14 06:32:36 vortex unix: WARNING: out of processes
> Dec 14 06:32:40 vortex last message repeated 8 times
> Dec 14 06:32:40 vortex unix: NOTICE: out of per-user processes for uid 214
> Dec 14 06:32:40 vortex unix: WARNING: out of processes
> Dec 14 06:33:11 vortex last message repeated 46 times
> 
> Now uid 214 is the ldm, so it is the likely culprit.  This happened once
> before several months ago.  At that time I recompiled ldm with just the
> '-O' option, since I suspected that a target option I originally used in
> compiling ldm might have been the cause.  But it appears I was wrong about
> that.
> 
> Info:  SparcStation 10/712MP, 512MB, Solaris 5.7, ldm 5.1.2 .
> 
> I couldn't find anything on this in the ldm-support archives after a quick
> search, so I thought I'd ask if you had any ideas.  I suppose it could be
> a lot of things, pqact spawns tons of processes.
> 
> Tom
> ------------------------------------------------------------------------------
> Tom McDermott                           Email: address@hidden
> System Administrator                    Phone: (716) 395-5718
> Earth Sciences Dept.                    Fax: (716) 395-2416
> SUNY College at Brockport
> 
> ------- End of Forwarded Message

Hi Tom,

Yes, the ldm spawns a fair number of processes, but usually
not a huge number.  The max number of rpc.ldmd processes
that you should see is the number of request lines in your
ldmd.conf plus the number of allow lines in your ldmd.conf
plus one. Then there's probably a pqact, a pqbinstats, and
whatever else you've put in your ldmd.conf file and whatever
else pqact spawns.

This sounds similar to the situation where the ldm went
haywire and spawned billions and billions of rpc.ldmd
processes.  How many of the do you see if you do 'ps -ef |
grep rpc.ldmd'?

If this indeed what is happening, the few times I have seen
this it was due to a incompatible queue e.g. someone
installed a new version of the ldm that required a new queue
but forgot to actually make a new queue.  But I suspect the
same thing could happen on some systems with a corrupt
queue.  Did you or anyone else change anything in your
installation?  How long has your ldm been running?

If it's not runaway rpc.ldmd processes, do 'ps -ef | grep
ldm' and see how many ldm processes there are.  I would be
interested in seeing this.

Anyway, to get out of the current mess I would suggest
stopping the ldm a remaking the queue.  Then watch to see if
it happens again.

Please let me know what happens.

Anne
-- 
***************************************************
Anne Wilson                     UCAR Unidata Program            
address@hidden                 P.O. Box 3000
                                  Boulder, CO  80307
----------------------------------------------------
Unidata WWW server       http://www.unidata.ucar.edu/
****************************************************