Re: Limits to number of monitored servers?

On Tue, Apr 19, 2005 at 01:15:06PM -0500, Brodie, Kent wrote:
> Now on to my question:   Are there any known limitations on the number
> of monitored systems?   Currently, I have conserver monitoring 31
> servers across two different terminal servers, and I'm about to connect
> up another 30 or so.    I'm interested to see if anyone has experiences
> along the lines of very large installations.

the only real limitation is how powerful of a machine you have to run
conserver (memory and cpu, mainly).  at the company i'm at we have two
large (many consoles) conserver hosts (both netra t1s - one with 512 and
one with 1gig of ram).  conserver is using 140MB of VM for 1152 consoles
and 234M of VM for 1314 consoles.  now, why the large memory difference?
honestly, it's probably because of a bug in the memory management when
the server gets a HUP.  we use the method of refreshing conserver so
that we don't have to restart the servers an bump people off.  i might
have to try and track down that bug (memory usage per process seems very
"off").  hmm...or it's all in the way it handles client connections.
but i thought i'd cleaned up that stuff...yeah, it'll take some digging.

anyway, there's one particular situation for you.  over a thousand
consoles on a relatively low-end (these days) machine.  startup time
isn't great (takes a minute or so, if i remember right), but hey, it
only happens once in a blue moon.  as long as you have, say 250MB of ram
for conserver to use, you can pretty easly handle a thousand consoles.
heck, that might even cover the memory usage for any ssh processes, if
you're using ssh to connect to terminal servers.

hopefully that helps and gives folks some insight into what i, at least,
consider a "large" conserver install.  i'm sure (i hope?) others out
there could beat these numbers, but it's something to use as a
guideline.

Bryan