Frequently Asked Questions (FAQ)

How can I acknowledge EXPLOR in my publications?

By mentioning the use of EXPLOR in your communications:

• Computational resources were partly provided by the EXPLOR Mesocentre which is hosted by the University of Lorraine *

• High Performance Computing resources were partially supplied by the EXPLOR centre hosted by the University of Lorraine *

How do I report to EXPLOR my publications that have used its resources?

By sending information to explor-contact@univ-lorraine.fr.

How do I connect to EXPLOR?

You can access your EXPLOR user account by using SSH or X2GO. (-ssh; -x2go)

I can't connect, who should I contact?

In case of connection problem, please send an email with a description of your problem to explor-support@univ-lorraine.fr.

What is X2Go?

x2go is a software which allows you to remotely access your EXPLOR frontend graphically (desktop) through an ssh connection. It is available as a client for Windows, Linux and MacOS. For more information, you can consult the website http://wiki.x2go.org/doku.php/download:start.

How do I transfer my files to/from EXPLOR ?

There are two ways to transfer your data. They are described on next page.

Why is my access machine called `vm-XXX' ?

In order to ensure the confidentiality of users, virtual servers, projects and resources, user accounts and projects have been anonymized.

Do I have Internet access from EXPLOR?

From your work environment (access front end), you have access to the internet. On the other hand, the compute nodes cannot communicate with the outside of EXPLOR.

I am attached to two projects, why are my connection parameters are different?

Each project has its own (and anonymized) environment, accessible via a dedicated virtual machine (access front). The connection parameters are therefore different for each project.

I am attached to two projects, how can I transfer my data from one project to the other?

The procedures for transferring data between two projects are explained in next page.

How do I use the servers I have purchased since my project?

A instructions manual will be sent to you at the installation of your servers which will explain how to access your servers under SLURM.

How to submit a job?

From your environment, you will be able to use the SLURM job manager to submit your jobs. Please consult the documentation available at Job submission.

How to connect interactively to a compute node

By reserving the node through the salloc command. For example, if you want to request a node of the std partition for 1 hour, enter : salloc -N1 -p std -t 1:00:00 srun -pty bash

Which partition to choose?

Depending on the type of job and your resource requirements (CPU, memory, GPU, etc), you will be lead to prefer one partition or another. The set of partitions is described on the page: Available calculation resources.

In general:

• the nodes of the old hf partition (cne[01-16]) are reserved for sequential or slightly parallel works (1 to 8 cores maximum).
• the gpu partitions are suitable for jobs requiring GPUs.

In case of doubt, you can send an email to support at explor-support@univ-lorraine.fr.

What computation time should I specify in my script?

The maximum computing time that can be allocated to you depends on the partition and the number of nodes you choose.

See Available calculation resources.

We recommend that you specify your request as precisely as possible so that the SLURM manager can optimise the use of resources.

The job I have submitted did not start, it remains in the queue. Why is it still in the queue?

The squeue command will give you the status of your job (last column "REASON"). Your pending job may be in status :

• Resources: your job is going to start soon, SLURM is reserving the resource needed to run it.
• Priority: one or more other jobs are ahead of you in the queue
• QOSGrpJobsLimit: limits are associated to each partition, project or user according to the requested resources.

These limits are defined at next page.

Your job will change status when these limits are no longer reached.

The job I submitted is on standby, when will it start?

The --start option of the squeue command can give you an estimate of when SLURM assumes your job will start.

When I submit my job, I get the following error: Could not select QOS, please verify resources selected?

In your SLURM submission file, the parameters you specified (partition, number of nodes, computation time, etc.) are not correct.

You can refer to the Computational resources and Job Submission technical documentation to verify the compatibility of your settings.

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

Invalid qos specification means that you have requested a resource to which you do not have access.

For more information, you can visit the next page.

What is the module command used for?

The module command is used to load software, compilers, etc. into the user environment. For more information about using the module command, you can consult the next page.

Is python installed in EXPLOR?

Several versions of python are installed in EXPLOR. They can be accessed via the 'module' command:

• python2 via the module load anaconda/2 command.
• python3 via the module load anaconda/3 command
• specific versions optimized by Intel, via the commands module load python/versionX/intel.

My software needs GPU resources, how can I specify this requirement to SLURM?

In your submission script, you must first choose a partition with GPUs, then specify the number of GPUs required by using the -gres option of the sbatch command (e.g. --gres=gpu:2 for 2 GPUs request).