[collectd] collectd and large MTU

[Florian Forster]

Hi Ulf,

On Mon, Oct 08, 2007 at 09:45:42AM -0700, Ulf Zimmermann wrote:
> Does collectd handle changed MTU on Ethernet? We have some servers
> where we changed over the weekend the MTU from 1500 to 9000 and now
> our Ethernet charts do not match with the graphs I got via mrtg on the
> switch side.

I assume you're talking about the graphs collected with the `interface'
plugin, i. e. `if_octets-*.rrd'?

Under Linux these values are queried either by reading `/proc/net/dev'
or by using `getifaddrs', default is to read from `/proc/..'. The
counters there count the octets that were sent over the link - I don't
think the MTU should play any role in that at all..

You can use some one-line-shell-magic (or possibly `iptraf' or something
like that) to check if the counters in /proc work as expected and
provide the `right' values.. I certainly hope they don't work correctly
though, since I have no idea whatsoever what the problem in collectd
might be.

If the counters seem wrong, I'd ask the author of the network card
driver if he has an idea what might cause this. If the counters are
about one sixth of what you'd expect due to the MRTG graphs, he probably
assumed an MTU of 1500 somewhere in his code..

Oh, last but not least: Cisco has had some problems with counters in the
past.. This would mean that the MRTG graphs should look different now,
though. Have you `old' data to check which graphs look different?

Regards,
-octo
-- 
Florian octo Forster
Hacker in training
GnuPG: 0x91523C3D
http://verplant.org/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
Url : http://mailman.verplant.org/pipermail/collectd/attachments/20071008/ac32ecc4/attachment.pgp