Re: [HTCondor-users] Slot reuse issue

[Cole Bollig <cabollig@xxxxxxxx>]

Hi Vikrant,

If the Schedd still has a claim on the dynamic lot and the next priority job fits within the slot requirements then the dynamic slot will be reused for the next job. This is true in recent versions also. It would make sense that the job would report the same number of allocated resources. You could set CLAIM_WORKLIFE to zero or a smaller number in the EP configuration to prevent the reuse of the slots. This will make the system less efficient.

-Cole Bollig

---

From: HTCondor-users <htcondor-users-bounces@xxxxxxxxxxx> on behalf of Vikrant Aggarwal <ervikrant06@xxxxxxxxx>
Sent: Wednesday, April 2, 2025 3:56 PM
To: HTCondor-Users Mail List <htcondor-users@xxxxxxxxxxx>
Subject: [HTCondor-users] Slot reuse issue

 

Hello Experts,

I know we are using the old version of condor 9.0.17, I checked the release notes but couldn't find anything related to this behavior fixed in the latest version.

First job using dynamic slot completed around 2025-03-25T06:48:52.080113-04:00

This job used 12cpus and 98G memory.

Another job landed on this machine 2025-03-25T06:48:52.856904-04:00, requesting 1cpu and 8GB of memory. With ~8 GB memory per core on this machine. This job also used 12cpu and 98G of memory to run, basically it seems like it has reused the slot. 

We use MODIFY_REQUEST_EXPR_REQUESTMEMORY and MODIFY_REQUEST_EXPR_REQUESTCPUS on worker nodes. 

Doesn't look like an expected scenario. Has anyone encountered this issue before?

Thanks & Regards,

Vikrant Aggarwal