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

[Condor-users] Possibility to submit part of a job cluster?



Hi,

I've got a brief question. Is it possible to tell condor at submit time
only to run a certain part of the job cluster? Briefly browsing the docs
haven;t provided me with a good alternative.

Background:

My job cluster consists of around 100 queues in a single submit file and
with a total of about 10000 single jobs. After a smaller "crash"[1]
Condor "thought" all jobs were finished, but that was not the case.

Since roughly 7000 of these were already finished, I naturally wanted to
start only the remaining jobs. This time I submitted the full cluster
again, put it on hold and removed all jobs which were already run before
with condor_rm. Not only does this create quite a bit of stress on the
schedd, but it can also cumbersome to extract the jobs which already run
from the results.

Is there another - cleaner - way to do that?

Cheers

Carsten

[1] I tweaked the single jobs' control script and made all consecutively
started jobs fail within seconds :(