On Fri, Oct 14, 2005 at 09:37:53AM +0200, Dirk Kastens wrote:
"All configuration of disk thresholds, load limits, which
processes to monitor etc. is done on the Hobbit server,
NOT on the client. This is to allow centralized configuration
of the monitoring system."
I think, this is not a good idea. I told the administrators
of the client systems that they are responsible for what
BB will monitor. If they install a new program, they have
to add the process name to the bb-proctab file to check
if it's running. Now they have to inform me about what
should be monitored and I have to edit the hobbit-clients.cfg
file. That's a lot of work, apart from the fact that the
file will be growing very quickly.
It's a design decision I made. The distributed configuration
doesn't work for my setup; I am sure you can find people
who prefer each of the setups.
Wouldn't it be possible to still use local configuration
files that are located in the client/etc directory?
Not the way Hobbit is built currently. So if that is a
problem, stick with the BB client for now.
It wouldn't be impossible to support both models. What currently
happens is that the client data is sent to the Hobbit server;
here it is fed into the hobbitd_client module and converted into
cpu, disk, memory etc. status messages. These are then sent
back to the Hobbit server.
So the simplest solution would be to run the normal Hobbit client
on your servers, but instead of sending data directly to the
Hobbit server, it would be fed into a local hobbitd_client
running on the monitored server. This does require some small
code changes in the hobbitd_client module, but not very much.
Regards,
Henrik