[Orca-users] Re: newbie questions about monitoring multiple hosts...

[Paul Haldane]

On Thu, 4 May 2000, Steve Gilbert wrote:

> My plan is to run the orcallator on the target systems, and
> then use a perl script that calls rsync to pull the data files
> back inside our firewall to a central collector host which will
> run Orca, etc.

Sounds fine.  We do basically the same thing here but access the files
using NFS.  I suspect rsync or similar would probably be better as on some
of our (much) quieter servers the orca NFS traffic becomes noticeable.

> 1. With this model, do I even need to install Orca at all on the
>    target machines?  It seems that I should just be able to install
>    SE and run orcallator.se on them.  If this is the case, where
>    will my data files be written?  Under the prefix directory that
>    is specified in /etc/rc3.d/S99orcallator?

That's right - you only need orca on the machine that is actually
generating the graphs.

> 2. The data files that orcallator.se generates are only the percol-*
>    files, right?  The RRD files are only used by the orca script that
>    runs on the collector machine?

Yup.  The idea is that orca takes as input plain text files which are then
interpreted according to the definitions in the cfg file.  This means that
it's relatively easy to gather stats from other types of machines (you
write something that works in a similar way to orcallator and produces a
text file, then add clauses to the orca.cfg to interpret that file).

> 3. By the same token, the orcallator.cfg file is also only used by
>    orca on the collector?  The target machiens should have no use
>    for that file?

Right again.

Paul

-- 
Paul Haldane
Computing Service
University of Newcastle