[HTCondor-users] Whole machine job with partitionable slots

[Vikrant Aggarwal <ervikrant06@xxxxxxxxx>]

Hello Experts,

I am exploring the possibility of running whole machine job with normal jobs. It's partitionable slots setup with almost homogeneous nodes.

I read some old HTcondor discussions [1], [2]. It looks like that I can use defrag to free the machine for scheduling the whole machine job but after reading discussion [2] (reply fromÂMichael Pelletier)Âit seems like that without defrag also we can insert the logic for detecting the whole machine jobs waiting in queue to adjust the STARTD _expression_ so that we can wait for completion of existing jobs, stop accepting new normal jobs and only allow to run whole machine job. Can anyone please point me to an example of achieving it?

We are already using scheduler splitting which should fill the depth first.Â

[1] https://www-auth.cs.wisc.edu/lists/htcondor-users/2014-February/msg00002.shtml
[2] https://www-auth.cs.wisc.edu/lists/htcondor-users/2017-January/msg00047.shtml


Regards,
Vikrant