On Jan 13, 2011, at 3:44 PM, Lawrence Strydom wrote:

> I understand that something is adding data and logically the backup should grow. What I don't understand is why the entire file has to be backed up if only a few bytes of data has changed. It is mainly outlook.pst files and MSSQL databse files that cuase these large backups. Some of these files are several GB.

Because Bacula is file-based (as are most other backup systems), and not, say, block-based (like, e.g., Norton Ghost), "many messages in a single file" mailbox formats like PST and mbox will tend to play havoc with backups, because even adding a single message to your mailbox will cause the whole mailbox to be backed up (i.e., the new message plus all the old messages in there).  That's one reason why Apple changed over to maildir-like message storage in their mail client when they introduced their Time Machine backup system---it is much friendlier to backups, as a new message only causes the file containing the new message to be backed up in a subsequent incremental backup.

You'll only get close to the sort of behaviour you want (i.e., only the changed data in the file is backed up) if and when Bacula gains some measure of deduplication support.  (Maybe not even then, depending upon how it decides to do it.)

> My understanding of an incremental backup is that only changed data is backed up. It seems that at the moment my Bacula is doing differential backups, ie backing up the entire file if the timestamp has changed, even though I have configured it for incremental.

If the last modified timestamp changes since the previous incremental backup then Bacula will assume the file has changed and include it in the incremental---even if the file data has not changed.

You can enable Accurate backups and check other attributes (such as MD5 checksums of the file) if you want Bacula to take more care in only backing up files that have truly changed.  This will slow down the backup speed, though.

Cheers,

Paul.