per-partition account resource limits

Discussion:

Nathan Harper

2014-09-30 13:00:36 UTC

Hi,

We're in the process of migrating away from our ageing SGE implementation
to SLURM. Although we've replicated almost all the functionality, one
thing I've not managed to achieve is per account resource limits on a
partition. There are times when I can't rely upon fairshare and want to
set harder limits. eg:

I hav partition1 with 10 nodes. GroupA can use all 10 nodes. GroupB can
only use 6 nodes, and GroupC only 4.

Any suggestions about how I might acheive this? With SGE I would set the
resource limit to something like:

limit users GroupA queues partition1 to slots=160
limit users GroupB queues partition1 to slots=96
limit users GroupC queues partition1 to slots=64
--
*Nathan Harper* // IT Systems Architect

*e: * nathan.harper-0eSKoRi+iP310XsdtD+***@public.gmane.org // *t: * 0117 906 1104 // *m: * 07875 510891 //
*w: * www.cfms.org.uk <http://www.cfms.org.uk%22> // [image: Linkedin grey
icon scaled] <http://uk.linkedin.com/pub/nathan-harper/21/696/b81>
CFMS Services Ltd // Bristol & Bath Science Park // Dirac Crescent // Emersons
Green // Bristol // BS16 7FR

[image: 4.2 CFMS_Artwork_RGB] <http://www.cfms.org.uk>

------------------------------
CFMS Services Ltd is registered in England and Wales No 05742022 - a
subsidiary of CFMS Ltd
CFMS Services Ltd registered office // Victoria House // 51 Victoria Street
// Bristol // BS1 6AD

Trey Dockendorf

2014-09-30 15:16:39 UTC

Permalink

QOS. Something like

PartitionName=partition1
AllowQOS=partition1-groupA,partition1-groupB,partition1-groupC

Then create QOS in sacctmgr with GrpCPUs or GrpJobs equal to "slots" values
from SGE. Not exactly the same but has the same effect. If you don't want
users to have to use the --qos argument when submitting, can use job_submit
plugin to assign the right QOS based on partition and their default account
(group).

- Trey

=============================

Trey Dockendorf
Systems Analyst I
Texas A&M University
Academy for Advanced Telecommunications and Learning Technologies
Phone: (979)458-2396

Post by Nathan Harper
Hi,
We're in the process of migrating away from our ageing SGE implementation
to SLURM. Although we've replicated almost all the functionality, one
thing I've not managed to achieve is per account resource limits on a
partition. There are times when I can't rely upon fairshare and want to
I hav partition1 with 10 nodes. GroupA can use all 10 nodes. GroupB can
only use 6 nodes, and GroupC only 4.
Any suggestions about how I might acheive this? With SGE I would set the
limit users GroupA queues partition1 to slots=160
limit users GroupB queues partition1 to slots=96
limit users GroupC queues partition1 to slots=64
--
*Nathan Harper* // IT Systems Architect
Linkedin grey icon scaled]
<http://uk.linkedin.com/pub/nathan-harper/21/696/b81>
CFMS Services Ltd // Bristol & Bath Science Park // Dirac Crescent // Emersons
Green // Bristol // BS16 7FR
[image: 4.2 CFMS_Artwork_RGB] <http://www.cfms.org.uk>
------------------------------
CFMS Services Ltd is registered in England and Wales No 05742022 - a
subsidiary of CFMS Ltd
CFMS Services Ltd registered office // Victoria House // 51 Victoria
Street // Bristol // BS1 6AD

Nathan Harper

2014-10-01 12:22:43 UTC

Permalink

Thank you for the response. I'd had a look at using QOS, but I had hoped
to avoid having to change the way that users do things (like having to use
the --qos argument). I've had a brief look at the job_submit plugin, but
I'm not sure where to start. Is there any thorough examples/documentation
out there?
--
*Nathan Harper* // IT Systems Architect

*e: * nathan.harper-0eSKoRi+iP310XsdtD+***@public.gmane.org // *t: * 0117 906 1104 // *m: * 07875 510891 //
*w: * www.cfms.org.uk <http://www.cfms.org.uk%22> // [image: Linkedin grey
icon scaled] <http://uk.linkedin.com/pub/nathan-harper/21/696/b81>
CFMS Services Ltd // Bristol & Bath Science Park // Dirac Crescent // Emersons
Green // Bristol // BS16 7FR

[image: 4.2 CFMS_Artwork_RGB] <http://www.cfms.org.uk>

------------------------------
CFMS Services Ltd is registered in England and Wales No 05742022 - a
subsidiary of CFMS Ltd
CFMS Services Ltd registered office // Victoria House // 51 Victoria Street
// Bristol // BS1 6AD

Post by Trey Dockendorf
QOS. Something like
PartitionName=partition1
AllowQOS=partition1-groupA,partition1-groupB,partition1-groupC
Then create QOS in sacctmgr with GrpCPUs or GrpJobs equal to "slots"
values from SGE. Not exactly the same but has the same effect. If you
don't want users to have to use the --qos argument when submitting, can use
job_submit plugin to assign the right QOS based on partition and their
default account (group).
- Trey
=============================
Trey Dockendorf
Systems Analyst I
Texas A&M University
Academy for Advanced Telecommunications and Learning Technologies
Phone: (979)458-2396

j***@public.gmane.org

2014-10-01 12:25:34 UTC

Permalink

see:
http://slurm.schedmd.com/job_submit_plugins.html

Post by Nathan Harper
Thank you for the response. I'd had a look at using QOS, but I had hoped
to avoid having to change the way that users do things (like having to use
the --qos argument). I've had a brief look at the job_submit plugin, but
I'm not sure where to start. Is there any thorough examples/documentation
out there?
--
*Nathan Harper* // IT Systems Architect
07875 510891 //
*w: * www.cfms.org.uk <http://www.cfms.org.uk%22> // [image: Linkedin grey
icon scaled] <http://uk.linkedin.com/pub/nathan-harper/21/696/b81>
CFMS Services Ltd // Bristol & Bath Science Park // Dirac Crescent // Emersons
Green // Bristol // BS16 7FR
[image: 4.2 CFMS_Artwork_RGB] <http://www.cfms.org.uk>
------------------------------
CFMS Services Ltd is registered in England and Wales No 05742022 - a
subsidiary of CFMS Ltd
CFMS Services Ltd registered office // Victoria House // 51 Victoria Street
// Bristol // BS1 6AD

--
Morris "Moe" Jette
CTO, SchedMD LLC