Re: [HTCondor-users] Complex policies to kill jobs?

[Greg Thain <gthain@xxxxxxxxxxx>]

On 08/23/2017 09:26 AM, Jose Caballero wrote:
> Hi,
>
>
> I am not sure this is even possible. Unclear to me from the documentation.
>
> Let's say that I have, for example, 10 running jobs and 10 idle jobs.
> And I want to kill 15 of them, with the following policies:
> -- first, as many idle as possible (in this case, the 10 of them)
> -- then, the rest (5) from the running ones, sorted by the time they
> started running, killing younger first.
>
> Is there a way to express this type of policies, either with the
> command line or -preferable- with the python library?

Do you want this as a one-time command, or as an ongoing policy?  If the 
former, condor_rm doesn't have  a rank or limit option, which might be a 
nice feature request, but as a shell one-liner, a quick hack that 
doesn't deal with held jobs or other edge cases:

condor_q -nobatch -af clusterid procid jobsstatus JobCurrentStartTime | 
sort -n -k 3,3r -k 4,4 | awk '{print $1 "." $2}' | tail -15 | xargs 
condor_rm

should do the trick.

-greg