If accidentally one or more files are lost from a home-directory (H:-disk) or a /vol-directory, probably the can be restored. On most C&CZ-servers every morning a "snapshot" is taken, on some servers even a few times each day. If one contacts C&CZ system administration the same day, most of the time we can easily restore the situation of that morning with this snapshot. In other cases we have to revert to the backup-tapes to restore the files. To be present on backup-tapes a file must of course be older than one day.
For home-PCs or other PCs that are not managed by C&CZ, the UCI TSM taperobot might be a good solution. Each employee can store up to 20GB on that robot without having to pay usage charges.
Different independent backup schemes are used:
- Morning snapshots to disk.
- Daily backups to tape.
- Weekly backups to tape.
- Yearly backups to tape.
Morning snapshots to disk
Many servers take a "snapshot" of their disks at 07:00 hours Monday through Friday, and on some Linux servers at 13:00 hours as well. With this snapshot C&CZ system administration can see the exact situation of that morning, which makes an "undelete" easy. If a user wants to be able to do this himself, C&CZ can create on each H:-disk where possible a shortcut/symbolic-link/bind-mount ".snapshot", which makes it possible for the user to look at the snapshot himself from a Windows-PC (or any Samba client).
From a Unix/Linux machine one can attach that H:-disk through:
or as root through:
mount -t cifs //$server/$user /$mountpoint -o user=$user
The regular servers for the home-directories of students and staff are pluri, bundle, pile, and goudsmit.
We cannot easily share the snapshot filesystem through NFS and an NFS-server does not follow 'symbolic links', that's why we have to do it as described above. The name of the shortcut/symbolic-link is chosen such that Windows shows it as a "hidden" file, in order not to confuse users. At this moment it is a pity that we cannot easily have snapshots on the home-server "servius" (Medical and Biophysics). Because we have snapshots, the old "Incremental backup to disk", that was hard to maintain and thought to be of less use than snapshots, was phased out.
Daily backup to tape
All server disk-partitions except temporary space like /tmp, /var/tmp, /scratch, are backed-up nightly on a number of LTO4-tapes. Because of the time it takes to do a backup, most of the times this is a so-called incremental dump, which means that only changes w.r.t. the latest full backups are written. These backups are started each evening at 21:00 hours. The backup-tapes are preserved for about five to six weeks.
A restore request can be made when the missing files are not in the incremental backups on disk
Weekly and yearly backup to tape
Every weekend a subset of all server disk partitions are completely written to a couple of LTO4-tapes. Every partition is written to tape roughly once every four weeks. These weekly backup-tapes are preserved for one year. The yearly backups are performed in December and are preserved for at least five years.