Re: [HTCondor-users] Force all jobs by a user on same execute node

[gagan tiwari <gagan.tiwari@xxxxxxxxxxxxxxxxxx>]

But that will be manual. We need it to be handled by condor.

On Fri, Jan 5, 2024 at 12:31âPM Carsten Aulbert <carsten.aulbert@xxxxxxxxxx> wrote:

Hi Gagan,

On 1/5/24 07:54, gagan tiwari wrote:
> Hi Guys,
>         ÂWe have a requirementÂthat all the jobs submitted by
> a user should work on the same execution node.
>
> ForÂexample, if a user submits 10 jobs and then all of these 10 jobs
> should run on the same node. If 10 slots aren't availableÂon any node in
> the pool , then all jobs should go to idle.


knowing nothing about the jobs themselves, wouldn't it be possible to
spawn all 10 jobs via a wrapper and start all 10 jobs via the wrapper
after requiring, say, 10 CPU cores on a single machine?

That way, they would only launch on a single node and only if there was
enough room in the (dynamic) slot.

Cheers

Carsten

--
Dr. Carsten Aulbert, Max Planck Institute for Gravitational Physics,
CallinstraÃe 38, 30167 Hannover, Germany, Phone +49 511 762 17185