Re: [HTCondor-users] Prioritizing jobs using the RANK expression on the execute point (EP) side

[Arshad Ahmad <aahmad@xxxxxxxx>]

Hi,

I followed your suggestion and set CLAIM_WORKLIFE = 0 on the execute node. I then submitted:

• Non-preferred jobs with varying requirements → ended up in two separate autoclusters

• Preferred jobs → also in separate autoclusters

I queued many preferred and non-preferred jobs, but they started running mostly in the order submitted, not giving visible preference to the higher-Rank jobs.

I can confirm that:

• Rank is being evaluated (preferred jobs show 1000, non-preferred 0)

• Despite this, Rank doesn’t seem to influence job start order in this setup

This seems similar to the behavior you mentioned regarding CLAIM_WORKLIFE. Even with it set to 0, the scheduler isn’t honoring Rank for job dispatch ― possibly due to how claims are reused or how the schedd schedules jobs in this particular node setup.

I’ll continue testing, but we may need to investigate if additional constraints or schedd settings are required to enforce Rank-based preference when multiple jobs are waiting on the same nodes.

Do you have any further suggestions on how to enforce preference for jobs from this group (the owners of these machines)? Please let me know if you need any specific logs or additional information to investigate.

Thank you.

-Arshad

---

From: Kühn, Max (SCC) <max.fischer@xxxxxxx>
Sent: Monday, September 15, 2025 6:13 AM
To: HTCondor-Users Mail List <htcondor-users@xxxxxxxxxxx>
Cc: Arshad Ahmad <aahmad@xxxxxxxx>
Subject: Re: [HTCondor-users] Prioritizing jobs using the RANK _expression_ on the execute point (EP) side

 

[EXTERNAL] ? This message is from an external sender