[Perfparse-devel] serviceperf.log and sending a kill signal to nagios

[Yves M. <yme...@li...>]

Hi Flo, hi, everybody,

I'm taking an old mail from Flo from this list.
Who has already tested this ?
Garry, could you write that method in the documentation as an alternate m=
ethod to run
perfparse, please ?
More below...

> - xpdfile* configuration in nagios did not work for me, but you should
> note in README that it also works if you use

Flo: this is probably because you compiled nagios without --with-file-per=
fdata

>
> host_perfdata_command=3Dprocess-host-perfdata
> service_perfdata_command=3Dprocess-service-perfdata
>
> in nagios.cfg and
> define command{
>          command_name    process-host-perfdata
>          command_line    /usr/bin/printf "%b"
> "$TIMET$\t$HOSTNAME$\t$OUTPUT$\t$PERFDATA$\n" >>
> /usr/local/nagios/var/host-perfdata.out
>          }
> define command{
>          command_name    process-service-perfdata
>          command_line    /usr/bin/printf "%b"
> "$TIMET$\t$HOSTNAME$\t$SERVICEDESC$\t$OUTPUT$\t$SERVICESTATE$\t$PERFDAT=
A$\n" >>
> /usr/local/nagios/var/service-perfdata.out
>          }
>
> in misccommands.cfg
>
> Flo


I just tested it. It works. But it is not so easy to do. Here are the tra=
ps to avoid :
1/ recompile nagios with --with-default-perfdata instead of --with-file-p=
erfdata. Those
are exclusive and if you put both, it will not work like we want.

2/ Do what Flo describes above.

3/ You cannot remove that file without sending a signal to nagios to rest=
art it. I
wanted to prevent some problems with that, but in fact, this is a limitat=
ion. Things
will change in the future.

So what you do is a script:
#!/bin/sh
mv /usr/local/nagios/var/service-perfdata.out /tmp/service-perfdata.out
perfparse (where the file to scan is /tmp/service-perfdata.out)
rm /tmp/service-perfdata.out


In the future (not so far), I think I will not support to send a signal t=
o nagios to
reboot it. The 2 reasons are that it is a limitation that I would get rid=
 of, and
because some users have problems with it.
So the new way to run perfparse will be that one. And of course, perfpars=
e will be able
to mv and rm the file if you want to delete it. You won't need the script=
.
In the future, perfparse will also be able to read from the command line,=
 allowing the
use of a pipe instead of a temporary file.
In the future(*), perfparse will also be able to write a compressed and r=
otated log with
what it parsed (without the need that nagios writes it somewhere)

Does anybody have a good reason to keep the current model with a signal s=
ent to nagios ?
Any comment, idea ?

If anybody wants to code some rrd or xml output right now, please contact=
 me :)

Thanks Flo and others for the ideas,
Yves

(*) that is already implemented in
http://ymettier.chez.tiscali.fr/perfparse-devel/perfparse-0.101.0ym3.tar.=
gz. But that
version does not connect to mysql yet. Work to do...

--=20
- Homepage    - http://ymettier.free.fr - http://www.logicacmg.com -
- GPG key     - http://ymettier.free.fr/gpg.txt                    -
- Maitretarot - http://www.nongnu.org/maitretarot/                 -
- GTKtalog    - http://www.nongnu.org/gtktalog/                    -