Reference

• Common Slurm Commands
• PBS to Slurm Migration
• Troubleshooting
• Security Installations
• PBS --> Slurm Converter
• HPC Price List

Common Slurm Commands

Quick reference for the most common Slurm commands on the TAU HPC cluster.

Submitting Jobs

Command	Description
sbatch job.sh	Submit a batch job script
srun --pty bash	Start an interactive session
sbatch --depend=afterok:JOBID job.sh	Submit job after another completes

Monitoring Jobs

Command	Description
squeue -u username	Your running and pending jobs
squeue	All jobs on the cluster
scontrol show job JOBID	Full details of a job
sacct -j JOBID --format=JobID,JobName,State,MaxRSS,Elapsed	Job accounting and memory usage
sattach JOBID	Attach to a running job's output

Managing Jobs

Command	Description
scancel JOBID	Cancel a specific job
scancel -u username	Cancel all your jobs

Cluster Information

Command	Description
sinfo	Partition and node status
scontrol show partition PARTITION	Partition details and limits
check_my_partitions	Your available partitions and accounts
features	Available node constraints/features

Environment Modules

Command	Description
module avail	List all available modules
module avail NAME	Search for a specific module
module spider NAME	Detailed module info including dependencies
module load NAME	Load a module
module list	List loaded modules
module unload NAME	Unload a module
module purge	Unload all modules

Common SBATCH Directives

Directive	Description
#SBATCH --job-name=NAME	Job name
#SBATCH --account=ACCOUNT	Account name
#SBATCH --partition=PARTITION	Partition/queue
#SBATCH --qos=QOS	Quality of Service
#SBATCH --time=HH:MM:SS	Max run time
#SBATCH --ntasks=N	Number of tasks
#SBATCH --nodes=N	Number of nodes
#SBATCH --cpus-per-task=N	CPU cores per task
#SBATCH --mem-per-cpu=NG	Memory per CPU
#SBATCH --mem=NG	Total memory
#SBATCH --gres=gpu:N	Number of GPUs
#SBATCH --constraint=FEATURE	Node constraint/feature
#SBATCH --array=1-N	Job array
#SBATCH --output=FILE_%j.out	Output file (%j = job ID)
#SBATCH --error=FILE_%j.err	Error file
#SBATCH --mail-user=EMAIL	Notification email
#SBATCH --mail-type=END,FAIL	When to notify

PBS to Slurm Migration

Quick reference for users migrating from the old PBS/Torque system to Slurm.

Command Equivalents

PBS/Torque	Slurm	Description
qsub job.sh	sbatch job.sh	Submit a job
qsub -I	srun --pty bash	Interactive session
qstat	squeue	View jobs
qstat -u username	squeue -u username	Your jobs
qdel JOBID	scancel JOBID	Cancel a job
pbsnodes	sinfo	Node status
qstat -f JOBID	scontrol show job JOBID	Job details

Directive Equivalents

PBS/Torque	Slurm	Description
#PBS -N name	#SBATCH --job-name=name	Job name
#PBS -q queue	#SBATCH --partition=partition	Queue/partition
#PBS -l nodes=1:ppn=8	#SBATCH --nodes=1 --cpus-per-task=8	Nodes and cores
#PBS -l mem=4gb	#SBATCH --mem=4G	Memory
#PBS -l walltime=02:00:00	#SBATCH --time=02:00:00	Wall time
#PBS -o output.log	#SBATCH --output=output.log	Output file
#PBS -e error.log	#SBATCH --error=error.log	Error file
#PBS -M email	#SBATCH --mail-user=email	Email
#PBS -m abe	#SBATCH --mail-type=BEGIN,END,FAIL	Mail events
#PBS -V	#SBATCH --export=ALL	Export environment
#PBS -t 1-10	#SBATCH --array=1-10	Job array

Environment Variables

PBS/Torque	Slurm	Description
$PBS_JOBID	$SLURM_JOB_ID	Job ID
$PBS_JOBNAME	$SLURM_JOB_NAME	Job name
$PBS_NODEFILE	$SLURM_JOB_NODELIST	Node list
$PBS_ARRAYID	$SLURM_ARRAY_TASK_ID	Array task ID
$PBS_NP	$SLURM_NTASKS	Number of tasks
$PBS_O_WORKDIR	$SLURM_SUBMIT_DIR	Submission directory

Key Differences

• Slurm requires --account and --qos — run check_my_partitions to find yours
• Slurm uses --cpus-per-task instead of ppn
• Memory in Slurm is per-CPU (--mem-per-cpu) or total (--mem)
• Slurm does not automatically change to the submission directory — add cd $SLURM_SUBMIT_DIR to your script if needed

Troubleshooting

Common errors and solutions for job submission and cluster usage.

Job Submission Errors

No partition specified

srun: error: Unable to allocate resources: No partition specified or system default partition

Always specify a partition. Run check_my_partitions to find yours.

Invalid account or partition

sbatch: error: Batch job submission failed: Invalid account or account/partition combination specified

Your account and partition combination is incorrect. Run check_my_partitions and make sure both match.

QOS not permitted

sbatch: error: Batch job submission failed: Job violates accounting/QOS policy

The QOS you specified doesn't match your account/partition. Run check_my_partitions to see valid combinations.

Job Failures

Out of Memory (OOM)

sacct -j JOBID -o JobID,JobName,State%20

JobID    JobName               State
-------- -------------------- --------------------
71       my_job        OUT_OF_MEMORY

Your job used more RAM than allocated. Resubmit with a higher --mem or --mem-per-cpu. To estimate needed memory:

sacct -j JOBID --format=JobID,JobName,MaxRSS,Elapsed

Timeout

Job state shows TIMEOUT — your job exceeded the time limit. Resubmit with a longer --time value.

Job stuck in Pending (PD)

Check the reason:

squeue -u username -o "%.18i %.9P %.8j %.8u %.2t %.10M %.6D %R"

Common reasons:

• Resources — cluster is busy, wait for nodes to free up
• QOSMaxCpuPerUserLimit — you've hit your CPU quota, wait for running jobs to finish
• InvalidQOS — wrong QOS, check check_my_partitions
• ReqNodeNotAvail — requested node is down, remove --nodelist constraint

NFS / Storage Issues

Job hangs or freezes on file operations

May indicate an NFS mount issue. Check if your home directory is accessible:

ls ~

If it hangs, contact HPC support — do not kill the job manually as it may cause further issues.

Disk quota exceeded

bash: cannot create temp file: Disk quota exceeded

Your home directory is full. Move large files to scratch space or contact HPC support for a quota increase.

Module Issues

Module not found

module avail MODULE_NAME

Check the exact module name. Use module spider MODULE_NAME for a broader search including partial matches.

Getting Help

If you can't resolve an issue, contact HPC support at hpc@tauex.tau.ac.il. Include:

• Your username
• Job ID
• The command you ran
• The full error message

Security Installations

Required security software for TAU workstations and servers.

NAC — Forescout

Network Access Control client required for connecting to the TAU network.

ForeScoutSecureConnector_64_visible_daemon.tar.gz

Installation

tar -zxvf ForeScoutSecureConnector_64_visible_daemon.tar.gz
cd secure_connector
./install.sh
systemctl start SecureConnector.service

EDR — CrowdStrike Falcon

Endpoint Detection and Response client.

• falcon-sensor_7.18.0-17106_amd64.deb — Ubuntu 22.04 / 24.04
• falcon-sensor_7.39.0-19203_amd64.deb — Ubuntu 24.04
• falcon-sensor-7.17.0-17005.el9.x86_64.rpm — Rocky / RHEL

Installation — Ubuntu

dpkg -i falcon-sensor_7.18.0-17106_amd64.deb
systemctl restart falcon-sensor.service

Installation — Rocky / RHEL

rpm -ivh falcon-sensor_7.17.0-17005.el9.x86_64.rpm
systemctl restart falcon-sensor.service

Registration (new installations only)

/opt/CrowdStrike/falconctl -s --cid=<cid-code>

To obtain the CID code, contact infosec@tauex.tau.ac.il.

PBS --> Slurm Converter

loading

HPC Price List