[slurm-users] Held jobs age priority accrual
Cyrus Proctor
cproctor at tacc.utexas.edu
Sat Oct 6 10:21:41 MDT 2018
Hello,
We've recently made the transition from version 17.11.5 up to 18.08.0.
Anecdotally, we think we're seeing a change in behavior regarding the
priority of held (user or admin) jobs. For discussion, take the example
where a user submits a job, it waits for a day in the queues, the user
does a uhold on the job for another day and then releases the job. Upon
release, with our current configuration, I would expect that a job's
multifactor age priority to reflect 1 day of waiting in the queue, not 2
days. Do you all agree with that assessment?
Our current configuration is:
$ scontrol show config | grep -i prio
PriorityParameters = (null)
PriorityDecayHalfLife = 2-00:00:00
PriorityCalcPeriod = 00:05:00
PriorityFavorSmall = No
PriorityFlags =
PriorityMaxAge = 7-00:00:00
PriorityUsageResetPeriod = NONE
PriorityType = priority/multifactor
PriorityWeightAge = 7000
PriorityWeightFairShare = 3000
PriorityWeightJobSize = 5000
PriorityWeightPartition = 3000
PriorityWeightQOS = 5000
PriorityWeightTRES = (null)
PropagatePrioProcess = 0
As you can see, no PriorityFlags (e.g. ACCRUE_ALWAYS) are set.
So, my questions are:
1.) Is my assessment of how hold priority works accurate?
2.) Are there major changes between versions 17 and 18 that would affect
this behavior?
3.) What settings, if any, need to be updated to see the desired behavior?
Thanks for your insights,
Cyrus
PS There does appear to be active work around "AccrueTime" (scontrol)
and some other new features in 18 that have me worried.
More information about the slurm-users
mailing list