File Systems
This page provides an overview on the different file systems in use in the Magellano infrastructure.
Home Directory
User personal home directories can be found at the path /home/users/.
Home directories are available on every node in the Magellano infrastructure.
The path is stored in the HOME environment variable.
You are the owner of your home directory, and only you have read-write permissions on its content.
The HOME path is a symbolic link to /gpfs/ddn/users.
This is an area of a GPFS (now IBM Storage Scale) high-performance parallel file system.
The home directory quota is 200 GB.
Info about your quota usage is stored in the .quota_usage file:
$ cat $HOME/.quota_usage
Home quota used 74 GB of 209 GB (35.70%)
Note
At the moment it is not possible to increase the home quota.
Group Areas
Some directories are created to address group needs. They are available under /gpfs/ddn and, as the path suggests, are
part of a GPFS (now IBM Storage Scale) file system.
Usage and quota info are stored in the .disk_usage file. As an example, for the teorici area:
$ cat /gpfs/ddn/teorici/.disk_usage
Disk used 95914 GB of 104857 GB (91.47%)
Access to these directories is based on group membership. Quota is shared per group.
CVMFS Repositories
A CernVM-FS mount point is availble on all nodes at the /cmvfs path.
We use the private CVMFS repository software.pi.infn.it to distribute software and container images in our internal network.
This can be accessed at the path /cvmfs/software.pi.infn.it.
Inside you will find the following subdirectories:
Subdirectory |
Description |
|---|---|
|
Installation path of licensed software |
|
Supported containers for Slurm job runtimes |
|
Modulefiles that load pre-defined sets of packages |
|
Contains shared Python Virtual Environments |
|
Installation path of shared software compiled with Spack |
In general, you will not need to access these paths directly.
All software available in this repo can be loaded using environment-modules (see Environment Modules)
A selected list of CVMFS repositories for High-Energy Physics is already available on the nodes. If the repository you need is not available, you can ask for it to be added by issuing a ticket to our Service Desk. Also check the CernVM-FS Repository Monitor.
Scratch
Some clusters have local storage available on the compute nodes. This storage is used as a high-performance scratch area for HPC jobs.
A scratch directory is automatically created for each job before the start of your script.
You may find it at the following path: $SCRATCH_ROOT/$SLURM_JOB_USER/$SLURM_JOB_ID.
The SCRATCH_ROOT path depends on the configuration of the partition.
See HPC Clusters for info about specific partitions.