I would think an incremental backup of an mbox server would take longer and use up more disk space though. Move one message out of the 10 year old inbox into a 5 year old archive mbox file, and now you have both files being fully backed up again. Or, if you have something trying to generate diffs, you have this heavy process trying to make the diff files for the two folders. Maildir, one file disappears in one folder, one appears somewhere else, and the disk space used is the size of the message.
No, incremental backup of maildir is a pain, if you want your backup to be usable. Unless your incremental backup code is clever or you choose to remove files on every backup.
For example, I use rsync scripts to backup all of my data twice a day. Just about anything else is easy to backup (excepting database servers, where I do a dump and sync that), maildir is a pain.
Backup up mbox files with rsync is easy, only the changes get sent, most of the time the files are just being added to so just the new messages get transferred.
We you do it using maildir however, because the files get moved and renamed when you read them, things are less smooth. It means that you end up with multiple copies of messages on the backup server.
I could fix this I guess by making my rsync scripts remove deleted files on each sync, but I have it set not to so I can get things back when I delete them by accident.
So what you really need to incrementally backup maildir properly is either:
- a script that creates a snapshot of each backup using symlinks on the backup server (there is such a script out there)
- a tool that uses hashes to spot files that have moved/renamed and move them on the backup server
I really should be using a script that does the first of these options and then maildir would be less painful, but I've never yet go round to setting it up.
Previous Topic
Index
