Re: [HTCondor-users] Group memory quota

[Tom Downes <downes@xxxxxxx>]

David:

 

This is how the HTCondor manual describes slot weight. *It so happens* that it defaults to CPUs, but it could be anything you want. I believe it would apply to your case.

 

SlotWeight: This specifies the weight of the slot when calculating usage, computing fair shares, and enforcing group quotas. For example, claiming a slot with SlotWeight = 2 is equivalent to claiming two SlotWeight = 1 slots. See the description of SlotWeight on page 258.

 

Tom

 

From: HTCondor-users <htcondor-users-bounces@xxxxxxxxxxx> on behalf of David Cohen <cdavid@xxxxxxxxxxxxxxxxxxxxxx>
Reply-To: HTCondor-Users Mail List <htcondor-users@xxxxxxxxxxx>
Date: Thursday, December 13, 2018 at 12:52 PM
To: "htcondor-users@xxxxxxxxxxx" <htcondor-users@xxxxxxxxxxx>
Subject: Re: [HTCondor-users] Group memory quota

 

Thanks Tom,

But I'm looking for a hard quota that will cap a group the same way it is done for CPU:

GROUP_QUOTA_group_somegrp = 500

 

 

David

 

On Thu, Dec 13, 2018 at 7:59 PM Tom Downes <downes@xxxxxxx> wrote:

David:

 

My approach to this is to use SlotWeight to assign each slot a weight that is equal to the greater of CPUs or CPUs*<average RAM per core>. For me that # is 2GB.

 

SLOT_WEIGHT = ifThenElse(Cpus >= Memory/2048, Cpus, Memory/2048)

 

You might be able to substitute a Machine ClassAd for 2048 if you have a non-uniform RAM/core ratio.

User priorities and queue order then end up looking reasonable, though you canât help the fact that someone using up all the RAM on a node is still blocking use of idle CPUs on that node.

Tom

 

From: HTCondor-users <htcondor-users-bounces@xxxxxxxxxxx> on behalf of David Cohen <cdavid@xxxxxxxxxxxxxxxxxxxxxx>
Reply-To: HTCondor-Users Mail List <htcondor-users@xxxxxxxxxxx>
Date: Thursday, December 13, 2018 at 11:44 AM
To: "htcondor-users@xxxxxxxxxxx" <htcondor-users@xxxxxxxxxxx>
Subject: [HTCondor-users] Group memory quota

 

Hi,

Is there a group memory quota?

I went through the documentation trying to find a way to prevent a group from taking hold of all the cluster memory and practically denying the use of free CPU by others.

 

 

Cheers,

     David

_______________________________________________
HTCondor-users mailing list
To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx with a
subject: Unsubscribe
You can also unsubscribe by visiting
https://lists.cs.wisc.edu/mailman/listinfo/htcondor-users

The archives can be found at:
https://lists.cs.wisc.edu/archive/htcondor-users/

Attachment: smime.p7s
Description: S/MIME cryptographic signature