Possible memory leak in sysstat

4 views
Skip to first unread message

grob

unread,
Oct 7, 2009, 9:55:17 AM10/7/09
to kulua-l
I've got an Oracle database on a machine running Red Hat AS 4.5. Over
the past week it's started to become unresponsive around 1AM and the
past couple of nights the database has crashed. I created a cron job
to check what was running at the time of the crash and got the
following.

root 4615 0.0 0.0 57088 932 ? Ss Oct06 0:00 crond
root 19201 0.0 0.0 72012 1228 ? S Oct06 0:00 \_
crond
root 19202 0.0 0.0 5360 948 ? Ss Oct06 0:00 |
\_ /bin/sh /usr/lib64/sa/sa2 -A
root 19205 99.4 58.2 33558092 19073504 ? R Oct06 71:36
| \_ find /var/log/sa ( -name sar?? -o -name sa?? ) -mtime +7 -
exec rm -f {} ;

It looks like sysstat is causing the system to crash. It only happens
on this one box. I tried uninstalling it and reinstalling it (I'm
obviously a converted Windows admin), but it still crashed. I'm
running sysstat-5.0.5-14.rhel4. I know there are updates, but I
figured there might be some underlying problem instead since it only
happens on this one box.

Just wondering if anyone had some troubleshooting tricks that might
help me get to the cause of this. I think my next step is upgrading
sysstat.

thanks ahead of time,
Ben

David Hageman

unread,
Oct 7, 2009, 11:24:51 AM10/7/09
to kul...@googlegroups.com
What type of machine is this? Specs? Processor? Memory?

Are you analyzing the data from sysstat?

Do you have quite a bit of files in /var/log/sa?

Have you tried running sa2 command during the day?


--
========================================================
D. Hageman <dhag...@dracken.com>
Dracken Technology, Inc. http://www.dracken.com/
========================================================

Rudy, Jared

unread,
Oct 7, 2009, 11:29:22 AM10/7/09
to kul...@googlegroups.com
I would also log disk space usage in your cron job and make sure you're
not running out of space on any of your partitions during that time.

Jared Rudy
UNIX Administrator
St. Francis Health Center
1700 SW 7th
Topeka, KS 66606
785-295-7942

Steve Nordquist

unread,
Oct 7, 2009, 3:12:03 PM10/7/09
to kul...@googlegroups.com
Someone/thing got it to scan a circular dependence, perhaps a /dev or
/sys item if not another virtual directory?
What about turning on Ss logging? (Given a tamed groffwall is
unavailable on that system, I suppose.)
Reply all
Reply to author
Forward
0 new messages