Hi Chandana,
----- Original Message -----
> All,
>
> The nightly cron job to merge logs is failing on a new Centos 6.4 box.
> I ran /usr/libexec/pcp/bin/pmlogger_daily with the -t option, but the
> trace (attached) is silent on why it is failing.
Hmmm, relevant bit of the trace is this ...
Performance metrics from host corona-int.m4u.com.au
commencing Sat Jun 22 15:17:42.916 2013
ending Sun Jun 23 07:56:22.874 2013
Archive timezone: EST-10
PID for pmlogger: 11308
Input archives to be merged:
20130622.14.27
20130622.14.27-00
20130622.14.28
20130622.15.12
20130622.15.16
/usr/libexec/pcp/bin/pmlogger_merge: line 253: 27438 Killed
$cmd
pmlogger_merge: Directory: /var/log/pcp/pmlogger/corona
pmlogger_merge: Failed: pmlogextract 20130622.14.27 20130622.14.27-00
20130622.14.28 20130622.15.12 20130622.15.16 20130622
... which is indeed lacking details as to why pmlogextract failed.
If this machine is recording the mysql metrics, and you have the patch
you sent me earlier today running on your mysql server, I suspect it
may be the root cause. That patch changed the type and units of one of
the mysql metrics, which pmlogextract wont be able to reconcile.
You might get more detailed information from a manual pmlogger_merge
on those archives (see the final "Failed" line above) - the -VV option
can be used to ratchet up its verbosity.
If it turns out this is from the pmdamysql change, you'll want to add a
new pmlogrewrite(1) file to help yourself and others out - see example
of /var/lib/pcp/config/pmlogrewrite/linux_proc_migrate.conf - IIRC, it
will automatically fix this issue (on all affected servers) once that
is in place.
cheers.
--
Nathan
|