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

Re: [HTCondor-users] kill vs. load avg (was: scheduling delay)

On 11/02/2018 05:23 PM, Michael Pelletier wrote:
> Is the kill expression evaluated like periodic expressions, where it's only checked once every five minutes?

Since Greg replied off-list, I'll repeat it here:

PREEMPT is what stops the job,
KILL hard-kills jobs that are not PREEMPT'ing nicely.

It is not clear to me how many delays are involved in there.

What I'm trying to do is give a job time to ramp up before scheduling
another one to the same host. What the scheduler currently does is land
$NUM_SLOTS jobs on the same host at once. Causing $NUM_SLOTS file
transfers, followed by $NUM_SLOTS unzips, etc. That use pattern is not
what I want on this particular host.

Does anyone know if NUM_SLOTS require a full restart, or will
condor_reconfig do?

Dimitri Maziuk
BioMagResBank, UW-Madison -- http://www.bmrb.wisc.edu

Attachment: signature.asc
Description: OpenPGP digital signature