[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [HTCondor-users] giving priority to a job already queued
Thanks for all the help. I was aware of condor_now, but I explicitly didn't want to remove a running job (my poor users...), I wanted to run this job next, once a slot would become available. (In practice, that would mean waiting about 2 min.) We recently switched from SGE as a batch system, so I was looking for the equivalent of qalter [jobid].Â The job in question wasÂ submitted as a debug user which tends not to have many jobs in the queue, so for this condor_userprio is probably the way forward. But for cases where a user has a lot of jobs in the queue, I would first have to re-order these jobs internally using condor_prio and useÂ then condor_userprio to give that user priority.
I guess what I really need is a "condor_inamoment" as a milder version of "condor_now".
On 6/18/2020 5:01 AM, htcondor-users@xxxxxxxxxxx wrote:
> Hi All,
> I have a job in the queue that I would really like to run as soon as possible.
> Is there an easy way to tell condor to run this particular job next ? (I'm the administrator, not an impatient user.)
I assume that all the slots on your pool are already busy, else this queued job would soon start running....
So I thus assume what you want to do is run this job instead of some other job that is currently running.
Administrators have the power to do this! :)Â Specifically, users specified in the condor_config with QUEUE_SUPER_USERS
have the power... by default user root and user condor are QUEUE_SUPER_USERS.
The solution is to use the condor_now tool.Â On your submit host (where the schedd runs), su to root (or to any other
QUEUE_SUPER_USER), and then run condor_now with the job_id of the important job you want to run now, and the job_id you
want to vacate and give the resources to your important job.Â ÂDo "man condor_now" for details, or see
Hope this helps,