[drraw-users] drraw "cluster"?

[Eric Law]

Joel Krauska,
What you might be really asking is for loose integration across multiple
disparate RRD installations?

I'm using cricket and have long surpassed the ability to collect and store
data using single hosts.
I've even been pushing IO limits on my shared storage.

Among the possible solutions (aside from just adding more/better/faster
hardware), are:
* reduce IO load on RRD updates (ie: keep more in memory for longer periods,
or optimize for updates like the rrd daemon alluded to would do).
* integrate disparate RRD respositories at the UI level  (the closest
solution applicable to the problem mentioned in the original email).

Imagine a drraw URL call that would export the data in csv/xml, much like
RRD XPORT does today....
Then imagine the drraw UI that would 'fetch' data (via the export feature)
from a _remote_ drraw UI and merge results together in a single display.
maybe the fetch is done via perl sockets/LWP within the CGI, or maybe the
fetch is done via AJAX at the browser.....

It could happen... it could work... it could exist before the RRD Daemon
becomes a reality (unless Tobi is up to something I'm not aware of ;-) )

Anyone want to work on such a beast?

-Eric

On 9/12/07, David Nolan <vitroth+ at cmu.edu> wrote:
>
> On 9/12/07, Joel Krauska <jkrauska at gmail.com> wrote:
> > It's perhaps possible to get around this using a shared filesystem for
> rrds, but it would
> > be nice to think about how drraw itself could accomplish this....
> >
> > Anyone else have this problem/need?
> >
>
> Absolutely, but I don't think this is a problem that drraw should try
> to solve.  I'd like to see a more generalized solution for distributed
> RRD collection & graphing.  I have got more devices/hosts I want to
> graph at my primary location then I can collect with a single machine,
> let alone graph...   I'm doing the shared filesystem via NFS for
> graphing.  Collection on two hosts, exported to a third for graphing.
> And I shudder every time I think about it.
>
> Tobias Oetiker (author of RRD) has been looking at adding an rrd
> daemon which handles the file IO, buffering and smoothing it, maybe
> even caching it for read access.  Once that happens I think
> distributed RRD access should be fairly easy.
>
> On the other hand...  Great, why don't you add that and send us a
> patch.  :)
>
> -David
> _______________________________________________
> drraw-users mailing list
> drraw-users at taranis.org
> http://web.taranis.org/mailman/listinfo/drraw-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://web.taranis.org/pipermail/drraw-users/attachments/20070912/bc3fd0c9/attachment.html