Re: [HTCondor-users] Slot weights

[Jeff Templon <templon@xxxxxxxxx>]

Hi,

this bit that I stole from Michael Pelletier on the HTCondor list (2017 IIRC) goes a very long way to achieving what I want, with standard HTCondor functionality:

# this is stolen from Michael Pelletier
# off the HTCondor-users list

SLOT_WEIGHT = Cpus + ifThenElse(TotalSlotMemory > (TotalMemory / TotalCpus * Cpus), (TotalSlotMemory - (TotalMemory / TotalCpus * Cpus))/(TotalMemory / TotalCpus), 0)

The âTotalMemory / TotalCpus * Cpusâ looks so wrong to me, but rearranging it to âCpus* TotalMemory / TotalCpusâ did not work for some reason, and despite the funky ordering, it does what I want it to do, as far as I can tell. The idea is that âTotalMemory / TotalCpusâ is the per-core memory available, so if you use more than this (per core), you get charged extra in the slot weight.  I am not sure yet, but I think this also has the effect of pushing large memory jobs towards machines that have more memory per core (correct?). In any case, it does have the desired effect in how many slots are allocated to whom.

JT


> On 21 Dec 2025, at 22:27, Jeff Templon <templon@xxxxxxxxx> wrote:
> 
> Digging quite deep in the docs, I did find this one mention:
> 
> SLOT_WEIGHT An expression that specifies a slotâs weight, used as a multiplier the condor_negotiator daemon during matchmaking to assess user usage of a slot, which affects user priority. Defaults to Cpus.
> 
> I think this should also be in the main section on priorities, with a statement of exactly how it affects the user priority.
> 
> JT
> 
> 
>> On 21 Dec 2025, at 22:06, Jeff Templon <templon@xxxxxxxxx> wrote:
>> 
>> Actually, itâs not clear whether the slot weight affects the priority.  It does affect the weighted in use values and total usage values printed by condor_userprio, but the documentation does not say anything about the effect of slot weight on priority, even though that effect would be The Right Thing To Have.
>> 
>> Gurus?
>> 
>> JT
>> 
>> 
>>> On 21 Dec 2025, at 21:59, Jeff Templon <templon@xxxxxxxxx> wrote:
>>> 
>>> Hi Christoph.
>>> 
>>> I donât think that has the intended effect, since it does not affect the job priority, no? The slot weight does affect the job priority as far as I can tell, which is what I want.
>>> 
>>> I found some gem on the list archives that I am trying out in test mode now.  More later!
>>> 
>>> JT
>>> 
>>> 
>>>> On 21 Dec 2025, at 20:50, Beyer, Christoph <christoph.beyer@xxxxxxx> wrote:
>>>> 
>>>> Hi Jeff,
>>>> 
>>>> I think for steering a ranking impression in the job would be the apropriate mean ... 
>>>> 
>>>> 
>>>> best
>>>> christoph
>>>> 
>>>> -- 
>>>> Christoph Beyer
>>>> DESY Hamburg
>>>> IT-Department
>>>> 
>>>> Notkestr. 85
>>>> Building 02b, Room 009
>>>> 22607 Hamburg
>>>> 
>>>> phone:+49-(0)40-8998-2317
>>>> mail: christoph.beyer@xxxxxxx
>>>> 
>>>> ----- UrsprÃngliche Mail -----
>>>> Von: "Jeff Templon" <templon@xxxxxxxxx>
>>>> An: "HTCondor-Users Mail List" <htcondor-users@xxxxxxxxxxx>
>>>> Gesendet: Sonntag, 21. Dezember 2025 16:55:51
>>>> Betreff: [HTCondor-users] Slot weights
>>>> 
>>>> Hi,
>>>> 
>>>> I am looking to use the SLOT_WEIGHT both to steer high-memory jobs to high-memory machines, and as well weight the usage of high memory jobs more than those of low memory, to make the distribution of slots to users more fair.
>>>> 
>>>> HOWEVER:
>>>> 
>>>> Enable use of the condor_negotiator-side resource consumption policy, allocating the job-requested number of cores to the dynamic slot, and use SLOT_WEIGHT to assess the user usage that will affect user priority by the number of cores allocated. Note that the only attributes valid within the SLOT_WEIGHT expression are Cpus, Memory, and disk. This must the set to the same value on all machines in the pool.
>>>> 
>>>> Really: the same value on all machines?  Why?  Or do I read this wrong? 
>>>> 
>>>> JT
>>>> 
>>>> 
>>>> _______________________________________________
>>>> HTCondor-users mailing list
>>>> To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx with a
>>>> subject: Unsubscribe
>>>> 
>>>> The archives can be found at: https://www-auth.cs.wisc.edu/lists/htcondor-users/
>>>> 
>>>> _______________________________________________
>>>> HTCondor-users mailing list
>>>> To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx with a
>>>> subject: Unsubscribe
>>>> 
>>>> The archives can be found at: https://www-auth.cs.wisc.edu/lists/htcondor-users/
>>> 
>> 
>