[rrd-developers] Accelerator Daemon

[Bernard Li]

Hi Florian:

On Mon, Jun 30, 2008 at 1:52 AM, Florian Forster
<rrdtool at nospam.verplant.org> wrote:

> Yes, that's probably right. I didn't expect anyone with such a huge
> number of RRD files to generate graphs periodically/statically instead
> of on demand/dynamically.. Is anyone with a big number of files doing
> that? If so: Why?

I am not sure if Ganglia fits your description, but I will describe
what it does.

A picture speaks a thousand words, the following URL demonstrates how
Ganglia uses rrdtool for graphing:

http://monitor.millennium.berkeley.edu/

The rrds are stored in /var/lib/ganglia/rrds and are partitioned by
the name of the cluster and by hostname.  We collect 30+ metrics which
are stored as separate .rrd files for each host and if you have a host
called "server1" in a cluster "cluster1", the directory hierarchy
would look like:

/var/lib/ganglia/rrds/
                             cluster1/
                                         server1/
                                                     boottime.rrd
                                                     bytes_in.rrd
                                                     cpu_user.rrd
                                                     ...

As you can see, as we monitor more hosts, more rrd files are generated.

While not all rrd files are used for graphing all the time, they are
still being updated frequently.

Hope this description is good enough -- if not, please feel free to
ask any questions.

Cheers,

Bernard