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

Re: [HTCondor-users] JOB_IS_FINISHED_* config options not removing jobs with delay



Hi TJ

Maybe you meant OnExitHold ? Simply toggling OnExitRemoveto to False will set it to Idle again and make it run again (IIUC the doc).

As an additional data point, we are using LeaveJobInQueue to keep jobs in the queue (and hence visible via condor_q) for some time w/o changing the job status, e.g.


LeaveJobInQueue = ifThenElse((JobStatus=?=4 || JobStatus=?=3) && (time() - EnteredCurrentStatus < 30 * 60*60), true, false)


Of course that needs to be done in the jobs submission, it is not a schedd config.

Maybe Mary wants something that can be enforced in the schedd w/o users action ? A JobTransform ?


Stefano


On 11/10/2024 16:02, John M Knoeller via HTCondor-users wrote:
These config knobs are meant to throttle the rate at which the Schedd will spend time removing jobs from the queue, not to provide a way to delay removal.   

So the delay value is measured from the time the first job is ready to be removed, there is not a separate timer for each job.  

Also using these throttling knobs will affect condor_rm as well as job completion, which is probably not what you want. 

The way to have jobs remain in the queue after they are completed is to use the OnExitRemove _expression_ of the job.  And to have them removed at some time in the future you can use SYSTEM_PERIODIC_REMOVE, or the PeriodicRemove _expression_ of the job.

-tj