BackupUpOnUnreliableLink
From RdiffBackupWiki
This is a solution we set up to do rdiff-backup on an unreliable wan connection.
Rdiff backup is transactional on a rdiff-backup session. That is to say, if a backup fails in the middle, on the next attempt, it will regress to the last valid state (the last backup that finished succesfully) and then start over. This is the desired behavior for this kind of software and that is a great feature of rdiff-backup.
However, in case you have to backup on an unreliable connection, you may end up never getting even one successful backup completly finished. Indeed since on every rdiff-backup session, the session will first start by regressing to the initial state...
We had this issue on an unreliable ISDN/ADSL links, and had to find a workaround. The workaround we are now using is coupling a local rdiff-backup with a remote rsync synchronisation. The local rdiff-backup session is fast because it is local and is processed on a reliable link (same host or same lan).
The rsync, use with the --partial option is a very robust way of transfering data over an unreliable link. We wrote a small script to restart rsync as long as it has not finished the synchronisation.
The system-backup.py script is available in different versions:
