Hi Lloyd,

The GrpCPUMins limit is a limit on the number of CPU minutes that an 
account can consume before they're stopped and unable to launch any more 
jobs. Once this happens their usage is decayed (at a rate of 
PriorityDecayHalfLife) or their usage is reset (according to 
PriorityUsageResetPeriod) in order to allow jobs to run against that 
account again.

The GrpCPURunMins limit on the other hand is used to limit the total 
number of CPU minutes that an account can have running at any moment in 
time.

So for example, if you were using hard limits that didn't decay - so you 
effectively gave accounts a certain amount of compute time (a certain 
number of CPU minutes) that they had to use within some period, say a 
quarter you would set:
PriorityDecayHalfLife=0
PriorityUsageResetPeriod=QUARTERLY
in your slurm.conf

and with sacctmgr you would allocate CPU minutes to accounts by setting 
the GrpCPuMins limit.

And, if you at the same time wanted to prevent users from having more 
than say 50,000 CPU hours (3M CPU minutes) of jobs running at the same 
time you could set GrpCPURunMins to 3000000.

That's my understanding anyway.

Hope it helps!
Mark


On 20/02/13 06:21, Lloyd Brown wrote:
>
> Can someone help me understand the difference between the GrpCPUMins and
> GrpCPURunMins limits?  I'm struggling to understand the difference,
> based on the descriptions in the sacctmgr manpage
> (http://www.schedmd.com/slurmdocs/sacctmgr.html).
>
>
>
> Thanks,
>
>

Reply via email to