Frequently Asked Questions (FAQ)

General

Question 1: How should I acknowledge EXPLOR in my publications?

Mention the use of EXPLOR in your communications:

* The computing resources were partially provided by the EXPLOR mesocentre hosted by the Université de Lorraine.

* High Performance Computing resources were partially provided by the EXPLOR centre hosted by the Université de Lorraine.

Question 2: How do I report publications that used EXPLOR resources?

Send information to explor-contact@univ-lorraine.fr.

Connection

Question 3: How do I connect to EXPLOR?

You can connect to your EXPLOR user account using SSH or X2GO (–ssh; –x2go).

Question 4: I can't connect, who should I contact?

For connection issues, send an email describing the problem to explor-support@univ-lorraine.fr.

Question 5: What is X2Go?

X2Go is software that provides graphical (desktop) remote access to your EXPLOR access node over SSH. Client versions are available for Windows, Linux and macOS. See http://wiki.x2go.org/doku.php/download:start for more information.

Question 6: How do I transfer files to/from EXPLOR?

Two options are available for transferring data. They are detailed on the following page.

Question 7: Why is my access machine named vm-XXX?

To guarantee user, server, project and resource confidentiality, user accounts and projects are anonymized.

Question 8: Do I have internet access from EXPLOR?

From your working environment (access node) you have internet access. Compute nodes, however, cannot communicate outside EXPLOR.

Question 9: I'm attached to two projects, why are my connection settings different?

Each project has its own anonymized environment accessible via a dedicated virtual machine (access node). Connection settings therefore differ per project.

Question 10: I'm attached to two projects, how can I transfer data between them?

Procedures for transferring data between projects are explained on the following page.

Question 11: How do I use servers I purchased from my project?

An operating procedure will be provided when your servers are installed explaining how to access them under SLURM.

Resource usage

Question 12: How do I submit a job?

From your environment, use the SLURM job scheduler to submit jobs. See the examples at Example submission scripts.

Question 13: How do I get an interactive session on a compute node?

Request a node reservation via salloc.
Example (request 1 node on partition std for 1 hour):
salloc -N1 -p std -t 1:00:00 srun --pty bash

Question 14: Which partition should I choose?

Choose a partition according to job type and resource needs (CPU, memory, GPU, etc). See the list at: Table of associations.

In general:

• nodes from the old hf partition (cne[01-16]) are reserved for sequential or low-parallelism jobs (1 to 8 cores max).
• gpu partitions are suitable for jobs requiring GPUs.

If in doubt, contact support at explor-support@univ-lorraine.fr.

Question 15: What runtime should I request in my script?

The maximum runtime depends on the partition and number of nodes chosen. See Available computing resources.

We recommend estimating runtime as accurately as possible to help SLURM optimize resource usage.

Question 16: My submitted job hasn't started and is queued. Why?

squeue shows job status (last column "REASON"). A queued job may be in status:

• Resources: SLURM is reserving the resources and the job will start soon.
• Priority: other jobs are ahead of yours in the queue.
• QOSGrpJobsLimit: limits apply per partition, project or user depending on requested resources.

These limits are defined at Limitations on job duration and resources.

Your job will change status when those limits allow it.

Question 17: My queued job — when will it start?

The --start option of squeue can give an estimate of when SLURM expects the job to start.

Question 18: On job submission I get: Could not select QOS, please verify resources selected ?

Your SLURM submission parameters (partition, nodes, time, etc.) are incorrect. Check the technical documentation: Computing resources and Job submission to verify compatibility.

Question 19: sbatch: error: Batch job submission failed: Invalid qos specification ?

Invalid qos specification means you requested a resource you don't have access to.

See Limitations on job duration and resources for details.

Question 20: What is the module command for?

module loads software, compilers, etc., into the user environment. See usage at this page.

Question 21: Is Python installed on EXPLOR?

Several Python versions are available via module:

• python2: module load anaconda/2
• python3: module load anaconda/3
• Intel-optimized specific versions: module load python/versionX/intel

Question 22: My software needs GPU resources — how do I request them in SLURM?

Choose a partition that provides GPUs, then request GPUs with the --gres sbatch option (e.g.: --gres=gpu:2 to request 2 GPUs).