On Wed, 15 Nov 2006, Martin Simmons wrote: > Incremental backups use the ctime as well as the mtime by default, so it > depends on whether mv changes the ctime (some file systems do, some don't).
rsync processes mirror both the ctime and mtime by default. This raises a problem if files older than the last backup are mirrored with the rsync process. > If Bacula does detect the moved file then there is also a problem with > restoring from incrementals, because it will create two copies of the file, > with the old and the new names. > > There is currently no way around this, so regular full backups are advisable. Synthetic backups would solve both problems. This has been discussed as a way of handling backup/restore of large filesets. Basically the only _reliable_ method of doing backups is to compare a snapshot of the current disk's index structure with a snapshot from the previous backup and then backup all files which have altered or appeared, no matter how old they may look. Having made the snapshots for accurate backups, keeping them means that "point-in-time" restores are trivial and there's no issue of previously deleted files "coming back to life"... It's a win-win situation. It also has major advantages if a backup is performed on an unmounted filesystem stub,(**) because the next backup will catch files which have changed, even if older than the last one. To my knowledge, almost every backup method out there relies on mtime/ctime stamps and is susceptable to the same problems with files with old stamps not being backed up. (**) This is _exactly_ what happened in september, resulting in loss of about a week's software development when a full restore was required recently.... AB ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys - and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
