Re: [HTCondor-users] peaceful DEFRAG_SCHEDULE?

[Carsten Aulbert <carsten.aulbert@xxxxxxxxxx>]

Hi

On 02.09.21 09:26, Beyer, Christoph wrote:
> you might come up with a ranking expression for longer running jobs to prefer hosts that have the longest time to live which presumably should cluster these jobs on a couple of machines ?

as we usually have many more jobs waiting for nodes become available, 
there is usually not much choice by RANK where to run. Also, I fear this 
would not work as the jobs do not advertise their presumed run times 
(e.g. because it is hard or impossible to estimate as it may very well 
be input data dependent).

Unfortunately, many of these jobs are also pieced together in ways that 
seems to make check-pointing too hard to implement :(.

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