[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[HTCondor-users] question on scheduler max jobs running

Hello HTC developers,

I am one of Fermilab glideinwms(GWMS) developer team.

One behavior of GWMS is, it stops using a scheduler when the sum of 
TotalRunningJobs and TotalSchedulerJobsRunning from that scheduler
is near MaxJobsRunning from the same scheduler.
(We are aware that the definition of MaxJobsRunning has changed since HTC 8.3.2)

We are trying to modify this behavior of GWMS such that GWMS can make this decision
based only on TotalRunningJobs, 
i.e. we want to ignore TotalSchedulerJobsRunning when GWMS makes this kind of decision
because scheduler universe jobs  (e.g. dagman jobs) are not handled by GWMS.

In the current 8.2.X release where MaxJobsRunning includes all universe jobs(except grid univ.),
do you provide any other scheduler parameter or scheduler classad that applies only to Vanilla universe jobs?

One very crude idea that occurred to me was defining a new limit inside GWMS code like
MaxJobsRunning-Vanilla- * X
where X can be a portion of vanilla running jobs in the total running jobs.

HyunWoo KIM
Fermilab Scientific Computing Division