[collectd] RandomTimeout question

[Jie Zhang]

Thank you so much for your suggestions! We are trying to use a binary rpm
instead of a source rpm for our product, I am sorry that I typed the rpm
name wrong. We don't have a lot of time to build the rpm from source and
write the init scripts.  Maybe it's a good idea to build it from a source
rpm.

For the data collection, we are collecting the data from a matrix in which
there are 72 drives, we are worried that if all the check happen at the same
time, we will run out of bandwidth. We will make all the exec external
scripts 'forever loop'. Thanks for the suggestion!

I have another question: we are collecting the data every 15 minutes and
will use each PDP without consilidating them, which means that we only need
the AVERAGE RRAs (MAX, MIN and AVERAGE should be the same as my
understanding), I am wondering if there is a way to configure the rrdtool to
save only the AVERAGE RRAs in the rrdfile?

Thanks!

Jie

2009/12/10 XANi <xani666 at gmail.com>

>  Dnia 2009-12-10, czw o godzinie 14:52 -0500, Jie Zhang pisze:
>
>
>
> I was trying to configure the rrdtool plugin to spread out writes to rrd
> files by setting the "RandomTimeout" value in rrdtool plugin configuration.
> But I got a messaging saying "Plugin `rrdtool' did not register for value
> `RandomTimeout'." when collectd were started. I did some more research and
>  it seems that the RandomTimeout patch might be added after the rpm package
> I am using was built.  I am using the latest collectd rpm I can found for
> CentOS,
> which is collectd-4.7.2-1.el5.rf.src.rpm and was built on Wed 29 Jul 2009.
>
> If the above is true , could anyone tell me where to find the patch and how
> to apply the patch to my rpm ( or if it's not a patch,
> How can I add this functionality to the collectd rpm I am using)?
>
>  Simplest way would be getting patch from git (git clone repo then use
> frontend like gitg to search for patch which applies that function).
> But, as you're gonna to compile collectd anyway u might just get newest
> 4.8.x source from page (patch for randomtimeout is only in =>4.8 tree afaik)
>
>  Otherwise, does anyone know what's wrong with my configuration? Here is
> the configuration of rrdtool:
> LoadPlugin rrdtool
> <Plugin rrdtool>
>         DataDir "/usr/var/lib/collectd/rrd"
>         RRARows      2880
>         RRATimespan  259200
>         CacheTimeout 900
>         CacheFlush   3600
>         RandomTimeout 10
> </Plugin>
>
> it's ok, tho in my experience RandomTimeout=CacheFlush/5-10 works best
> (will spread out much faster after restart of collectd)
>
>
> On the other side, I also want the collection of the data be staggered so
> that the monitoring won't all occur in bursts.
> Does collectd have this capability?
>
> You mean that all checks wont be done at same time ? Do some of your checks
> lag your servers so much u need it ? If those are "exec" external scripts u
> can make "loop forever" instead of "return value and exit" script, which
> usually eats bit less cpu and u can change interval to something different
> than default
>
>
>
> --
> Mariusz Gronczewski (XANi) <xani666 at gmail.com>
> GnuPG: 0xEA8ACE64http://devrandom.pl
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.verplant.org/pipermail/collectd/attachments/20091211/8eecaa78/attachment.htm