[HTCondor-users] peaceful DEFRAG_SCHEDULE?

[Carsten Aulbert <carsten.aulbert@xxxxxxxxxx>]

Hi all,

currently DEFRAG_SCHEDULE can be any one out of graceful, quick and fast.

However, in a pool with many jobs not doing checkpoints at all or 
ignoring SIGTERM altogether, I think it may be beneficial to really wait 
for the jobs to finish if it only affects very, very few nodes and one 
could accept the badput - just raising 
MaxJobRetirementTime/MachineMaxVacateTime would probably not do much as 
most jobs will still not intercept SIGTERM and still be killed via a 
graceful shutdown.

Would there be other consequences I currently do not see, is there 
already something in Condor I am overlooking or would this just be a 
small change in Condor to have this as a new feature?

Cheers

Carsten

--
Dr. Carsten Aulbert, Max Planck Institute for Gravitational Physics,
CallinstraÃe 38, 30167 Hannover, Germany, Phone +49 511 762 17185

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature