[collectd] cpu wait time on collectd server

[Florian Forster]

Hi Andi,

On Wed, Sep 12, 2007 at 12:31:23PM +0300, Andi Lõhmus wrote:
> Now when looking at the stats from the server, i see very large cpu
> wait times averaging about 70%. it must be waiting on network IO
> because disk write throughput is only ~1Mb/sec, which the disk can
> handle with ease. 

No, I'm pretty sure that's disk-IO. Take a look at this article which
describes in detail what the problem is:
<http://oss.oetiker.ch/rrdtool-trac/wiki/TuningRRD>

> Is such behaviour expected?

Yes. No..

> What can be done to optimize things?

Appart from the system-tuning-tipps in the above article you should
activate caching in the rrdtool plugin. The setting is `CacheTimeout'
and should be set to something like `300' or so.. Then 30 updates
(assuming the default interval of 10 seconds) will be written at once
which should reduce seeking- and read-ahead-overhead a lot..

Since I'm very interested in the performance of collectd it'd be nice if
you could report back if that setting helped to reduce your IO-wait and
by how much..

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/20070912/f27a2abb/attachment.pgp