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

Re: [HTCondor-users] jobs state switch from R to I in a cycle

On 10/2/2015 8:22 PM, Patricia Bittencourt wrote:

       I'm using condor 7.8.5 in a Rocks 6 cluster. And, I changed
condor configuration to give priority to "production" user jobs by
setting Rank = (Owner == "production"). I also reset some other settings
as the uid of the submitter didn' t look like it was being used during a
job execution.

       The issue now is the state of the jobs keep switching from R to
I, over and over again. But most of the time they are I. All those jobs
are merely a sleep for 2min and they never end as their state changes
all the time.

From the ShadowLog, it appears that the job 19.0 started running on machine nc09 but was immediately kicked off (evicted). Was job 19.0 a job with Owner == "production" ? If it was not and yet there were user production jobs idle in the queue, it would make sense for HTCondor to preempt job 19.0 in favor of running a production job. If 19.0 was a production job, then it would help if we could see the StartLog on the execute machine during this time frame to learn why HTCondor kicked off job 19.0.