[collectd] Plug-in wait time

Florian Forster octo at verplant.org
Mon Apr 20 16:36:34 CEST 2009


Hi Amit,

On Mon, Apr 06, 2009 at 12:35:02AM +0530, Amit Gupta wrote:
> It seems that the collectd plug-in's wait time to invoke the read
> callback function of any plugin keeps on increasing (doubles) if the
> plugin read function fails for some reason.

that is correct.

> Due to this behavior, apache plugin's read callback function (or any
> other server plugin) gets invoked after a long time if the apache
> server is shutdown for sometime and then becomes online. This
> apparently means that the apache statistics won't be collected by
> collectd for quite sometime even if the apache is online.

Right. We had to find a reasonable compromise between trying again and
again and leaving “dead” systems alone. Having this exponential back-off
means that the statistics will stay inactive for at most twice as long
as the system was down. I. e. if the Apache was unreachable for two
hours, you will have no statistics for at most four hours.

> Is there a way to configure the collectd (some kind of global option
> or a plugin specific option) to let it call the read callback function
> of a plugin after a fixed interval of time even if the read function
> fails (from the collectd's code, it doesn't seem possible though other
> than to hack the plugin to return a success value even if the read
> fails)

Such a value exists, but is currently hard-coded to 86400 seconds, i. e.
one day. Using a configurable value for this shouldn't be a lot of work.

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/20090420/de99b561/attachment.pgp 


More information about the collectd mailing list