[Orca-users] Better way to handle orca "instances"?

[Adam R Prato]

Hi there,

	I last looked at orca in 2000. I didn't set it up then because I didn't
like that there was a single orca page that grouped a set of hosts together 
(and if you had 100 hosts, this could get cumbersome). Fastforward to 2004,
and it seems orca is still the best thing out there for monitoring hosts 
(especially for solaris).

	Two weeks ago, I had some spare time, so I took a look again, and came
up with a way to cajole orca into logically organizing the groups the way I
desired. This seems to work pretty well:

- I modified start_orcallator to look for /etc/orca/group to determine a 
  subdir of orcallator to store the orcallator data files

- I created separate orcallator.cfg files for each group which has the paths:
   
base_dir                /prod/unix/orca/rrd/orcallator/@GROUP@
html_dir                /prod/unix/orca/html/@GROUP@
find_files              /prod/unix/orca/orcallator/@GROUP@/(.*)/(?:(?:orcallator)|(?:percol))-\d{4}-\d{2}-\d{2}(?:-\d{3,})?(?:\.(?:Z|gz|bz2))?

  I then use sed to replace @GROUP@ for the paths and @DESC@ for the html
  title.

- I run orca from cron every 30 minutes. I have about 5 groupings of hosts,
  and thus I run orca -v -once $group.cfg for each group. The total orca run
  takes about 15-20 minutes on a U60 with 1x360MHz USII processor and 2GB RAM.

- The (internal) clients write their orcallator data over NFS (though I have
  some external clients which I'll be setting up with rsync to pull the data).
  The filesystem for the collected data, rrd's, and html dirs are all local
  to the machine

Now for the questions.

- Is there a better way of managing groupings of hosts?

  I tried separate orcallator sources, with the find_files parameters specific
  to those groups and used redundant plot{} statements to generate the data,
  but I would still have a single huge html page with all of the hosts, which
  wasnt what I was looking for. 

- Is there *any* way to get orca to handle multiple configuration files so that
  it doesnt have to incur the startup costs for each of the N groups?

  Right now I think this is the biggest problem I'm having, likely due to the
  setup I have, the orca run takes so long to complete, and in that duration
  orcallator then accumulates so much data that the next run will take long, so
  on and soforth. I'm sure if I had a single config that this would run much
  more quickly, but at the sacrifice of having an orderly page of plots.

Any ideas or comments?

Adam