Thomas Guyot-Sionnest wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> If anyone's interested here's what ePN versus pure Perl plugins can
> looks like on the server's CPU load. This is on a dual proc
> (multi-threaded == 4 proc) server with hardware SCSI raid
> (battery-backed write-back) and 2GiB of RAM.
>
> The server is almost doing 1000 active check/minutes spread on nearly
> 100 hosts (Nagios 2.7). In the attached graph you can see the server
> load as I changed the settings of one Perl plugin that runs 28 times
> every minute. There's two drops:
>
> 1. At the very beginning I removed the "/usr/bin/perl " in the
> checkcommand definition. The ePN compilation was failing so the file was
> likely still being recompiled at every run. The drop in user time seems
> to be from the plug-in not running (but still being compiled).
>
> 2. After the week-end, the 2nd drop shows the drop when I fixed the
> plug-in compilation. At that time the plugin was running again but not
> being recompiled at each run, which caused a raise in user time and a
> nice drop in system time.
>
> This graph clearly shows that ePN is a must for any big system that rely
> heavily on Perl plugins. The difference you see here is caused by only
> 28 check out of nearly 1000.
>
Have you found a way around the memory leakage? Otherwise, I still believe
it's more hassle than it's worth, and effort would be better spent to cut
the number of fork()'s in half by having Nagios multiplex its checks.
Nice figures though.
--
Andreas Ericsson [email protected]
OP5 AB www.op5.se
Tel: +46 8-230225 Fax: +46 8-230231
This post was automatically imported from historical nagios-devel mailing list archives
Original poster: [email protected]