Re: [Condor-users] Submission of large numbers of jobs

[Ben Cotton <bcotton@xxxxxxxxxx>]

On Wed, Jul 11, 2012 at 4:21 PM, Brian Candler <B.Candler@xxxxxxxxx> wrote:
> On Wed, Jul 11, 2012 at 09:30:40AM +0200, Martin Kudlej wrote:
>> You can configure more than one scheduler.
>
> One thing which confuses me a bit. The info at
> http://research.cs.wisc.edu/condor/manual/v7.8/3_2Installation.html
> says:
>
> "For a machine that only submits jobs, there will be processes for:
> condor_master
> condor_schedd"
>
> This seems to say that every host which can submit jobs *must* maintain its
> own local job queue.  Is that correct?
>
Not exactly. That's how most sites (for certain values of "most". I'm
not sure anyone really has a good idea of what the "average"
installation looks like) are configured. A small number of submit
hosts serve the whole pool. It is possible, I believe, to use
condor_submit to submit a job to a remote schedd such that the machine
you're logged in to doesn't have to run a condor_schedd process. You
would use the -remote argument, along with the name of the remote
schedd.

> And does this in turn imply that condor_q talks to localhost, rather than to
> $(CONDOR_HOST) ?
>
> (A quick experiment suggests that condor_q is not affected by CONDOR_HOST,
> but condor_status is)
>
That is correct. CONDOR_HOST defines the central manager that the host
talks to (and thus the pool that it is in). condor_q will, by default,
check the schedd running on localhost. You can check another schedd
with: condor_q -name your.remote.server

On Wed, Jul 11, 2012 at 9:19 PM, Sarnath K - ERS, HCLTech
<k_sarnath@xxxxxxx> wrote:
> I just don't find a single architecture diagram that shows the different nodes, what they do etc..
> At least, a basic diagram would help understand things better..

Check out slide 4 of
http://research.cs.wisc.edu/condor/CondorWeek2012/presentations/adesmet-admin.pdf
for a diagram. Note you can combine multiple roles onto the same host.

-- 
Ben Cotton
Purdue University