SUMMARY: SAR (sa2) Seg Faults

Thanks to Kevin Buterbaugh, Mike Moehlman, Robert Alexander, and Darren Dunham for
their advice/counsel/suggestions.  I appreciate the time and effort.

I ended up finding the problem and solution on my own, though.  By default, the sa1
script is run once every 20 minutes and the sa2 script expects the data to be
collected at that time interval.  I had modified the times of the sa1 runs to once
every 15 minutes, to give me finer granularity on the data.  (This change was needed
due to the activity monitoring required on our CAD compute servers.)  sar was
occasionally choking on the input data when it tried to generate the report created
during the sa2 script run.

The fix was to remove the "-i 1200" from the cron entry for sa2.  According to the
sar man page, in the absence of the -i, it will use all available data.  That did the
trick.  

It's curious that the process ran successfully on some dates, but seg faulted on
other dates, all on the same box with the same parameters.  TTMYGH.

----Original Post:

Gurus,

To preface: I searched the list archives, sunsolve and the rest of the internet,
but
came up empty, so...

I'm seeing segmentation faults caused by the sa2 SAR process which runs on a cron
every evening.  Sometimes the process runs without problem and other times it seg
faults.  There doesn't seem to be a pattern as to architecture, time, activity on
the
machines, etc.  I'm seeing about 12 of these per night (= 10% of my network).  It's
not always the same machines every night.

On our network of about 120 Sun boxes (E280R, E250, E420, E3500, Ultra5, Ultra2,
Ultra1, SPARC20), we just upgraded almost all to Solaris 8 (07/01) with
8_Recommended
patches (11/05/2001).  In the process of the move to Sol8 (it was a clean install;
not an upgrade), I turned on SAR on all machines.  Used the standard sys cron and
uncommented the lines in /etc/init.d/perf to turn on SAR.

Is there a patch I missed?  This isn't a huge problem, but more an annoyance I want
to clear up.  Thanks in advance.  Will summarize.

-- 
Ed Alexander
Senior Systems Administrator
MEUS, Design Engineering Center - East
2635 Meridian Parkway
Durham North Carolina 27713
ed.alexander@msai.mea.com