Hi Jan-Frode,
On Sat, 2007-06-02 at 22:40 +0200, Jan-Frode Myklebust wrote:
> On Thu, May 31, 2007 at 03:38:31PM +1000, Nathan Scott wrote:
> >
> > I've switched the script over to have these now, and also added
> > the additional "very verbose" (-V -V) diagnostics that the
> > pmlogger_check script has - could you try out the attached
> > script, in place of your current /usr/share/pcp/bin/pmie_check?
>
> I didn't replace /usr/share/pcp/bin/pmie_check, but rather put
> your script in /etc/cron.hourly/pmie_check.sh. Unfortunately
> it also leaks out new instances for already running pmie's.
I found a bug in this earlier version of my patch - you may have
better results with the one in my git tree ("nathans" branch of
git://oss.sgi.com:8090/nathans/pcp.git).
The other thing I came across on our RHEL4 production servers
recently is that the tmpwatch(1) program is run daily from cron;
and it "cleaned up" some tmp files below /var/tmp for an agent
we use here. Since the pmie scripts also live in /var/tmp, its
also quite possible that it could eat the temp pmie state files.
We worked around this by adding the "-s" switch to the tmpwatch
invocation on /var/tmp (in /etc/cron.daily/tmpwatch).
cheers.
--
Nathan
|