[Orca-users] Orca using over 40% CPU on Server

[Liston Bias]

On Wed, 18 Jun 2003, Garrett, Matt M SITI-ITDIEEE wrote:

> Folks
>
> I have been running Orca Version 0.27 for over 6 moths now.
> It does produce very good detail and superb graphs
>
> However running on a Sun Solaris 8 system , Ultra 30 , UltraSPARC-II
> 296MHz CPU , 512Mb memory The Orca process seems to all ways use over
> 40% of the CPU The only reason it does not use 100% is other monitoring
> process (Nagios) need to run as well.
>
> At the moment I am only monitoring 53 clients.
> This is about to be stepped up to 300 clients.
>
> Is the CPU going to go through the roof and grind the poor monitoring
> server to a halt !
>
> The Orcas config file for base looks like
> base_dir                /local/data1/orcallator_column_data
>
> With all clients writing to the above via NFS , Note this is a local
> disk to the server , NFS mounted to the clients
>
> I had thought of getting Orca just to generate the RRD files and then
> once a day getting it to generate the images / html files but this kind
> of take's away the usefulness of nearline reports, and not sure if it
> would make any real difference.
>
> Any suggestions or idea's would be very helpful

Pegging your CPU is often a good thing.  In order to decrease your "update
time" on graphs, untilization that will need to be as high as possible.
I'm getting over 90% utilization on V480R used to run orca graphs on about
250 systems in order to keep the graph updates under 10 minutes.

You will probably get very large delays when you increase your clients
from 53 to 300.  Orca will probably have have a lot of trouble keeping up
with that many systems on an Ultra30.

I don't think you will gain much by running once a day.  I would just keep
it running nonstop and realize that you may have long delays between
graphs getting regenerated.  Let me/us know if you testing yields
different best case.  There was discussion at some point to rewrite
orcallator.se to generate RRD per-systems some time ago, but I don't think
anyone has put the time/effort into figuring this out.  If someone has,
that would be worth trying.

Splitting up your orca load amoung multi-processes or systems will also
help if you have the CPU's.

You may also look at eliminating graphs you don't need.  The less data
that orcallator has to process, the less work it will be doing.  This data
is specified in your cfg.  Many folks eliminate disk or network graphs.

- Liston