Hi Alfonso,

This is what fairshare points are supposed to help with and do mostly.  Still, 
It can take some time to get jobs started with big resource requirements at 
busy times even with high fairshare points.  There isn’t a perfect solution to 
get jobs started immediately from what I can tell unless you preempt/suspend 
jobs.  The problem with your solution is that you will have nodes sitting 
around idle I’d imagine.  Maybe instead some of the backfill parameters and 
other scheduling parameters could be tweaked.  Maybe it’s worth posting the 
relevant scheduling config parameters for some of the slurm pros to analyze?  
Maybe you could create a QOS with very high priority to help with this 
situation?  I have a debug QOS that has super high priority with limits: 32 
cpus (max node configuration), 30 minute wall time, 1 job at a time per user.  
This QOS helps greatly with testing … and isn’t abusable too easily - this 
likely isn’t a fit for your situation tho but maybe this thought could help.

Best,
Chris

—
Christopher Coffey
High-Performance Computing
Northern Arizona University
928-523-1167

> On Feb 20, 2015, at 8:35 AM, [email protected] wrote:
> 
> 
> I would strongly recommend against trying to change the select/cons_res 
> plugin for anything.
> 
> The job_submit plugin should be able to do what you want
> 
> Quoting "Pardo Diaz, Alfonso" <[email protected]>:
>> Hello,
>> 
>> I want to implement a new feature for SLURM, maybe a plugin. This new 
>> feature consist of some nodes reserved for new jobs in a busy environment. 
>> In other words, I have a cluster with "X" nodes, and �X-n� nodes are 
>> busy running jobs. If a user submits a job, only one job, I want the �n� 
>> reserved nodes run the new job. If the user submits more than one job, these 
>> jobs will wait in the normal queue.
>> 
>> What I want to get with this behavior is that occasional users' jobs don�t 
>> wait a long time for idles nodes.
>> 
>> I thought modifing the plugin �select/cons_res� will be a possibility. 
>> Is this correct?
>> 
>> 
>> Thanks!!!
>> 
>> 
>> Alfonso Pardo Diaz
>> System Administrator / Researcher
>> c/ Sola n� 1; 10200 Trujillo, ESPA�A
>> Tel: +34 927 65 93 17 Fax: +34 927 32 32 37
>> 
>> [CETA-Ciemat logo]<http://www.ceta-ciemat.es/>
>> 
>> ----------------------------
>> Confidencialidad:
>> Este mensaje y sus ficheros adjuntos se dirige exclusivamente a su 
>> destinatario y puede contener informaci�n privilegiada o confidencial. Si 
>> no es vd. el destinatario indicado, queda notificado de que la 
>> utilizaci�n, divulgaci�n y/o copia sin autorizaci�n est� prohibida 
>> en virtud de la legislaci�n vigente. Si ha recibido este mensaje por 
>> error, le rogamos que nos lo comunique inmediatamente respondiendo al 
>> mensaje y proceda a su destrucci�n.
>> 
>> Disclaimer:
>> This message and its attached files is intended exclusively for its 
>> recipients and may contain confidential information. If you received this 
>> e-mail in error you are hereby notified that any dissemination, copy or 
>> disclosure of this communication is strictly prohibited and may be unlawful. 
>> In this case, please notify us by a reply and delete this email and its 
>> contents immediately.
>> ----------------------------
> 
> 
> -- 
> Morris "Moe" Jette
> CTO, SchedMD LLC
> Commercial Slurm Development and Support

Reply via email to