Getting Started
Everything you need to access and begin using TAU HPC resources.
Access & Setup
How to connect to the HPC cluster and set up your environment.
Accessing the System
Requirements
- You must be a member of the power group
- Valid Tel Aviv University username and password
- If connecting from outside the TAU network: TAU VPN must be active
Login Node
All access to the cluster is via SSH through the login node:
slurmlogin.tau.ac.il
Your connection is automatically load-balanced across powerslurm-login, powerslurm-login2, and powerslurm-login3.
Connecting via SSH
ssh your_username@slurmlogin.tau.ac.il
With an SSH key:
ssh -i /path/to/your/private_key your_username@slurmlogin.tau.ac.il
Important
- Do not run compute jobs on the login node — use
sbatchorsrun - The login node is shared — heavy processes will be killed
Palo Alto VPN
TAU uses Palo Alto GlobalProtect VPN with two-factor authentication (Google Authenticator).
Required if connecting to the cluster from outside the TAU network.
Enrollment
- Go to https://mytau.tau.ac.il/GetResource.php and register your mobile phone
- Install Google Authenticator on your mobile device
- Scan the QR code provided during enrollment
Download
Download the appropriate version for your system:
Install
RHEL/Rocky/CentOS:
tar -xzf PanGPLinux-6.x.x-cx.tgz
yum localinstall GlobalProtect_UI_rpm-*.rpm
Debian/Ubuntu:
tar -xzf PanGPLinux-6.x.x-cx.tgz
dpkg -i GlobalProtect_UI_deb-*.deb
Configure
- Open the GlobalProtect client
- Enter gateway address: vpn.tau.ac.il
- Log in with your TAU credentials
- Enter the code from Google Authenticator when prompted
Troubleshooting: SSL Error on Ubuntu 22.04+
If you see an SSL error after connecting, apply this fix:
Open /usr/lib/ssl/openssl.cnf and add:
[openssl_init]
ssl_conf = ssl_sect
[ssl_sect]
system_default = system_default_sect
[system_default_sect]
Options = UnsafeLegacyRenegotiation
Restart the GlobalProtect app.
Environment
Managing software modules and storage on the cluster.
Environment Modules
Environment Modules let you dynamically load and unload software packages without conflicts between versions. Always load the modules your job needs before running it.
Common Commands
Finding Modules
# List all available modules
module avail
# Search for a specific module
module avail gcc
# Get detailed info including dependencies
module spider gcc/gcc-12.1.0
Loading and Unloading
# Load a module
module load gcc/gcc-12.1.0
# List currently loaded modules
module list
# Unload a specific module
module unload gcc/gcc-12.1.0
# Unload all modules
module purge
Inspecting a Module
# See what environment variables a module will set
module show gcc/gcc-12.1.0
In Job Scripts
Load modules inside your job script, after the #SBATCH directives:
#!/bin/bash
#SBATCH --job-name=my_job
#SBATCH ...
module purge
module load gcc/gcc-12.1.0
./my_program
Starting with module purge ensures your job is not affected by any modules loaded in your shell session.
Storage and Scratch
Home Directory
Every user has a personal home directory at /home/your_username. This is your default working directory when you log in.
- Backed up via Legato — see TAU backup info
- For purchasing additional storage, see storage pricing
- NetApp storage includes snapshots for file recovery
Scratch Partitions
Scratch partitions are shared, high-speed temporary storage available across the cluster:
/scratch100/scratch200/scratch300
Use scratch for intermediate files during a job run — not for long-term storage.
Local Scratch
Some compute nodes and workstations have a local /localscratch partition. This is node-local storage — faster than shared scratch but only accessible from that specific node.
If your job uses /localscratch, you must clean up after yourself. Add this to your job script:
export CACHEDIR=/localscratch/${USER}_${SLURM_JOB_ID}
mkdir -p $CACHEDIR
cleanup() {
rm -rf -- "$CACHEDIR" || true
}
trap cleanup EXIT INT TERM HUP
Important
- Scratch is not backed up — do not store anything you cannot afford to lose
- Clean up scratch files after your job completes
- Do not use scratch as a permanent storage location
First Steps
Your first steps on the cluster — finding your resources and submitting your first job.
Finding Your Account and Partition
Before submitting jobs, you need to know which account and partition you have access to. These are required parameters for every job submission.
Check Your Partitions
Run this command on the login node:
check_my_partitions
This lists all partitions and accounts you are authorized to use.
Key Concepts
- Account — your billing/group account (e.g.
public-users_v2). Required for all job submissions. - Partition — the queue your job runs in (e.g.
power-general-shared-pool). Determines which nodes are available and what resource limits apply. - QOS — Quality of Service, controls priority and limits (e.g.
public). Usually matches your partition.
Useful Commands
# View all partitions and their status
sinfo
# View partition details including limits
scontrol show partition power-general-shared-pool
# View your running and pending jobs
squeue -u your_username
Need Access?
If check_my_partitions returns nothing or you are missing a partition, contact the HPC team at hpc@tauex.tau.ac.il.
Submitting Your First Job
Once you have your account and partition from check_my_partitions, you are ready to submit your first job.
A Minimal Job Script
Create a file called first_job.sh:
#!/bin/bash
#SBATCH --job-name=first_job
#SBATCH --account=public-users_v2
#SBATCH --partition=power-general-shared-pool
#SBATCH --qos=public
#SBATCH --time=00:10:00
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=1
#SBATCH --mem-per-cpu=2G
#SBATCH --output=first_job_%j.out
#SBATCH --error=first_job_%j.err
echo "Hello from $(hostname)"
echo "Job ID: $SLURM_JOB_ID"
Submit It
sbatch first_job.sh
Slurm will return a job ID:
Submitted batch job 12345
Monitor It
# Check job status
squeue -u your_username
# View output once the job completes
cat first_job_12345.out
Job States
- PD — Pending, waiting for resources
- R — Running
- CG — Completing
- CD — Completed
- F — Failed
Next Steps
Once your first job runs successfully, see Running Jobs for arrays, GPU jobs, interactive sessions, and more.
See Also
- HPC Helper Toolkit — AI-powered tool to help with QOS configuration, job submission, and more
Useful Tools
External tools and resources to help with TAU HPC usage.
HPC Helper Toolkit
An AI-powered toolkit to help with common HPC tasks including QOS configuration, job submission, and more.