[collectd] network plugin questions

[Florian Forster]

Hi Oleg,

On Fri, Apr 18, 2008 at 07:17:51PM +0400, Oleg King wrote:
> 1. What is cache that controlled by CacheFlush setting?
> 
> OK, plugin caches time values and flushes time values, but what for
> it needs/uses theese values?

for loop detection, mostly. So if you have two instances, and both are
set to `forward', you will not end up with an endless loop. Also, the
`Forward' option makes use of this cache.

In general, you shouldn't need to worry about that setting. The default
of 30 minutes should be okay for pretty much everybody ;)

> 2. If I reboot server with RRD database, all information from clients
> will be lost until server will not up again. May be, it will be
> useful to implement 3rd, TCP type of connection with acknowledges of
> data reception and caching unsent info on clients side until server
> will be able to receive info again?

I've been thinking about some sort of redundancy protocol, so that
several servers can work together and replace one another. I haven't
actually coded anything in that direction yet, though. If you're
interested in implementing anything, we should definitely talk about our
ideas ;)

> 3. As I understand, clients sends information with time of event, but
> this is clients time of event. So we need to syncronise time on all
> servers to get right slices of database info?

Why? If the load at time 8am was `100', why write that value as if it
appeared at 2pm, just because it took us a while to process the value?
Anyway, synchronizing the time of your servers is always a good idea, so
that logfiles and administrative emails end up with a right time in
them..

> In one of my projects I calculate timediff between client and server
> and use this timediff to make all times syncronised. Will it be useful
> in collectd?

Ugh, time difference calculations are tricky. Very tricky, even. I'd
just install NTPd everywhere and be good..

> I will try to implement theese features anyway, because I need it, but
> may be I wrong somewhere?

Well, I use the time of the system collecting the data on purpose and
currently I don't see the point why this should be a bad thing. I see,
however, *a lot* of problems if you use the time of the system writing
the values to disk, because various levels of caching may be between
collection and storage of the values..

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/20080420/9b644255/attachment.pgp