Backups and snapshots on the Cluster filesystems¶

Below is an overview of the main backup and snapshots policies on UTHPC Cluster filesystems. If these policies are not suitable for your use/case, please contact support@hpc.ut.ee with a request.

UTHPC reserves the right to change and update these backup and snapshots policies as deemed necessary without prior notice.

Snapshots¶

Snapshots on a GPFS filesystem are read-only, point-in-time copy of the files in a fileset (larger subdivision of GPFS) at the moment of creation of snapshot. Snapshots retain the same access privileges as the master fileset and are user accessible. Snapshots are not a full backup, but rather a safeguard against accidental deletions or user error in data management.

The fileset in the snapshot is captured from a location that corresponds to the top-level directory of the file system. For example, the home fileset is currently linked to /gpfs/helios/home/ with the snapshots available from a subdirectory named .snapshots.

In general, snapshots are captured as follows:

Manual creation of snapshots is possible and is done before large-scale modifications to data by discretion of an UTHPC administrator.

Backups¶

UTHPC Cluster file systems are backed up weekly.

UTHPC Cluster file systems are backed up to tape storage. Backups are usually made file system wide, with the exception of the home fileset. This means that user home directories are not backed up. For any file backed up, up to 2 inactive copies exist along with an active copy. An inactive copy is a previous version of the file backed up with the active copy being the latest.

An inactive copy beyond the second one expires in 30 days after the backup process runs. For a "hot" file that has been deleted from the file system, the copy is kept for 60 days after the backup process has flagged the master file as deleted.