graphing system usage (was Re: [lug] forcing certain services to use eth1 instead of eth0)

Kevin Fenzi kevin at scrye.com
Wed Jan 4 15:44:16 MST 2006


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

>>>>> "John" == John Hernandez <John.Hernandez at noaa.gov> writes:

John> I see.  I wonder if something else in the IO subsystem was
John> overburdened, rather than the NIC itself?  1 x Gig-E is a LOT of
John> bandwidth.  Have you tried tracking (or graphing) the interface
John> utilization over time, such as with MRTG or similar?

I'll throw in a plug here for munin: 

http://munin.projects.linpro.no/

I maintain the munin packages for fedora-extras, so if you have a fc4
machine you should just be able to do: 

yum install munin munin-node

(The munin package is the central server/collector, munin-node runs on
each of the machines you want to monitor). 

Munin is nice because it auto-detects all the items on a particular
machine on install, so you have to do very little config to start
monitoring a ton of things. 

In this instance, munin might show you that your cpu is the bottleneck
or disk or irq's or whatever. 

kevin
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)
Comment: Processed by Mailcrypt 3.5.8 <http://mailcrypt.sourceforge.net/>

iD8DBQFDvE/C3imCezTjY0ERAhiWAJ40+pEd7LKoWgvMB/Ce3bKugK7LfgCfYwNk
mrwm9e8/itb3ixiDVEpayp0=
=PaJ3
-----END PGP SIGNATURE-----



More information about the LUG mailing list