Best Practices

This section gathers essential recommendations for the responsible and efficient use of GridUnesp resources. Following these practices contributes to the proper operation of the cluster for all users.

Sharing Access

Warning

Never share your password with third parties.

As a security measure, your account is personal and non-transferable. Members of the same project, advisors, or even the GridUnesp team must not have access to your password.

Risks of sharing:

  • Accidental deletion of files (cannot be undone)

  • Unauthorized modification of data

  • Submission of jobs in your name

  • Compromise of cluster security

How to collaborate safely:

If you need to share files with other users, use access permissions:

Adjusting permissions for collaboration
# Grant read permission to the group
chmod g+r shared_file.dat

# Grant read/execute permission to a directory
chmod g+rx shared_directory/

# Remove permissions for others (safe default)
chmod o-rwx shared_directory/

Tip

For collaboration on projects, ask your coordinator about creating group directories with appropriate permissions.

Requesting Support

Before requesting support, check the Frequently Asked Questions section - most questions have already been answered.

When sending an e-mail to ``support.ncc@unesp.br``, include:

  1. A clear description of the problem

  2. The location of the submission script (full path: /home/$USER/path/to/file/script.sh)

  3. The location of the input files (if applicable)

  4. Log files (slurm-JOBID.out)

  5. The command used to submit the job

  6. The number(s) of the job(s) in question (if any)

Important

The more information you provide, the faster and more accurate the support team’s response will be.

Data Maintenance

Backup Policy

Danger

GridUnesp does NOT have an automatic backup system.

In case of an unrecoverable failure of the storage system, all data will be lost with no possibility of recovery.

Recommendations:

  1. Transfer important results to your local computer

  2. Keep copies in at least two different locations

  3. Do not use the cluster as the only copy of critical data

  4. Keep only the files needed for processing on the cluster

Disk Space Management

Since resources are shared, unnecessary accumulation of data harms all users.

Check your space usage:

Analyzing usage in /home/
du -scm $HOME/$USER/* | sort -nr | head -20

Cleanup in /home/:

Removing unnecessary files
# Remove specific files
rm slurm-*.out obsolete_program

# Remove directories (be careful!)
rm -rf old_results/ temporary_tests/

Attention

The /store/ partition

Jobs that use multiple nodes or have the variables SHARED_FS="true" or LARGE_FILES="true" write to /store/. This partition also requires periodic cleanup:

Checking space in /store/
du -scm /store/$USER/* | sort -nr | head -20

# Remove directories of old jobs
rm -rf /store/$USER/2835861 /store/$USER/2835862

# OR remove everything (only if you are sure!)
# rm -rf /store/$USER/*

Danger

When storage reaches 100% usage, the entire system is compromised, affecting the jobs of all users.

Processing on the Access Server

Important

The access.grid.unesp.br server must NOT be used to run jobs.

Functions of the access server:

  • Entry point to your account

  • Preparation of scripts and files

  • Submission of jobs with sbatch

  • Light pre/post-processing tasks (fast and low memory)

Permitted exceptions:

  • Editing small files

  • Quick compilations

  • Organizing directories

  • Very brief tests

Danger

The GridUnesp team monitors the use of the access server. Heavy processing will be immediately canceled so as not to harm other users.

Best Practices Summary

✅ Always:

  • Use sbatch to submit jobs to the processing nodes

  • Keep backups of important data

  • Clean up temporary files regularly

  • Include complete information when requesting support

❌ Never:

  • Share your password

  • Run jobs on the access server

  • Use the cluster as your only storage location

  • Accumulate unnecessary data

See also