Full backups on our machines are done once a week. Files which have changed since the last full backup are dumped daily. (These dumps are called incrementals.) If a filesystem should fail, then by restoring a full dump and the most recent incremental on top of it, the filesystem can be restored to within a day of the failure.
Once the next full backup is done, the incrementals are obsolete. They are recycled for use as incremental tapes immediately.
In order to do a file restore, the operator needs to be able to determine which tape the file is likely to be on. It's possible that the file could be on any of the incrementals, the last four weekly fulls, or one of the monthly fulls.
For files on the NetApp (this now includes users home directories for DCS faculty, research, graduate, and undergrad clusters), backups are a little different. Snapshots (analagous to to backups) are done on the following schedule:
| Type | Scheduled | Number kept |
|---|---|---|
| hourly | Daily at 0400, 0800, 1200, 1600, 2000 | 10 |
| daily | Tuesday thru Sunday at 0000 | 12 |
| weekly | Monday at 0000 | 9 |
No snapshots or backups of files are available beyond the 9 week timeframe. (This is a new policy in response to the recently enacted law regarding public access to government records.)
Therefore, to request a file restore you should send a message to "operator@cs" specifying the following information for the file you need restored:
/fac/users/watrous/public_html/restore.htmlThis tells the operator which set of tapes the file would be in (for machines still backed up to tape) or on which volume the file existed (for files on the NetApp).