[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [HTCondor-users] problems with jobs requiring more then 2GB memory



The job is running out of memory because it is only requesting 2Gb of RAM but then using more than that. 

SLOT_TYPE_1_PARTITIONABLE=TRUE

Means that a slot with the amount of cpus and memory requested by the job will be created when AP decides to run that job, up to a maximum of 8 CPUs and 4 GB, because

SLOT_TYPE_1=cpus=8, memory=4096

To fix this, you need to change the request_memory of the job's submit file to request more memory

-tj


From: HTCondor-users <htcondor-users-bounces@xxxxxxxxxxx> on behalf of Mihai Ciubancan <ciubancan@xxxxxxxx>
Sent: Friday, May 30, 2025 2:28 AM
To: htcondor-users@xxxxxxxxxxx <htcondor-users@xxxxxxxxxxx>
Subject: [HTCondor-users] problems with jobs requiring more then 2GB memory

Hello,

I encounter issues with LHCb jobs ,which are requiring more than 2GB per
jobs. The jobs are failling with the following error:

LastHoldReason = "Error from reserved-LHCb2_5@xxxxxxxxxxxxxx: Job has
gone over cgroup memory limit of 2048 megabytes. Last measured usage:
2033 megabytes.  Consider resubmitting with a higher request_memory."

I have configure partionable slots:

CLAIM_WORKLIFE=3600
CONTINUE=TRUE
JOB_RENICE_INCREMENT=10
KILL=FALSE
NUM_SLOTS=4
NUM_SLOTS_TYPE_1=4
SLOT_TYPE_1_PARTITIONABLE=TRUE
SLOT_TYPE_1=cpus=8, memory=4096
SLOT_TYPE_1_START=Owner=="pillhcb01"
SLOT_TYPE_1_NAME_PREFIX=reserved-LHCb
PREEMPT=FALSE
RANK=0
SUSPEND=FALSE
SLOT_TYPE_1_CONSUMPTION_POLICY=False
CONSUMPTION_POLICY=False
CLAIM_PARTITIONABLE_LEFTOVERS=False

Also is enable cgroup policy:

BASE_CGROUP = /system.slice/condor.service
CGROUP_MEMORY_LIMIT_POLICY = soft
MAXJOBRETIREMENTTIME = $(HOUR) * 24 * 7
SYSTEM_PERIODIC_REMOVE =  ResidentSetSize > 3000*RequestMemory

If you have any suggestion will be highly appreciated!



Best,
Mihai

_______________________________________________
HTCondor-users mailing list
To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx with a
subject: Unsubscribe

Join us in June at Throughput Computing 25: https://urldefense.com/v3/__https://osg-htc.org/htc25__;!!Mak6IKo!K29kgDu3KqY-v0JvPE9cVXxO9hKbX4vVgC2pMuc85_5TCTwv4huZH_KU-ElZEvUc6BvAtLM_1S1Sk8MicXaY$

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