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

[HTCondor-users] round-robin broken?

Hi, folks.

I recently updated our cluster from Debian Jessie to Debian Stretch, and thus updated htcondor to 8.4.11 (the current version in Debian Stretch) from 8.4.8 (a previous stable snapshot from the Jessie repo at http://research.cs.wisc.edu/htcondor/debian/). It seems like a pretty minor update, but jobs are no longer being distributed round-robin across the machines in the cluster. If 100 jobs are submitted, we'd rather distribute the jobs across the machines evenly, instead of sequentially filling up each machine until all cores are used, and then moving along to the next machine.

I have "NEGOTIATOR_POST_JOB_RANK = -SlotID" in /etc/condor/config.d/01myconf (and "condor_config_val NEGOTIATOR_POST_JOB_RANK" gives the expected answer), but jobs are no longer being routed accordingly. I'm not seeing anything obvious in the logs. Any thoughts?

Grant Goodyear