On Sun, Dec 16, 2007 at 02:30:12PM -0800, David Brown wrote:
I appreciate the effort SGI has put into xfsdump, and that is why I'm using
XFS on my machines.
Ok, I completely take this XFS compliment back. I now have two
filesystems that fail to restore, utterly. It was unable to handle
the things that changed on an incremental backup and just refuses to
continue to restore.
This is quite sad. The _only_ backup program I've now encountered
that anything resembling reasonable behavior is GNU tar. At least
when it screws up, the failure is usually localized to one part of
the backup.
- dump/restore: complete failure to restore, even when dumping
snapshots.
- xfsdump: not sure exactly, but it looks like if the same inode
number is reused from a file to a directory then it won't
restore it.
- star: same kinds of problems.
The problem with all of these is that they try to defer the hard
part of figuring out what to do to restore. Most people don't test
restore very often (if ever) so they look like they are working,
when they are happily writing out backups that are not restorable.
Others that I've tried, including BRU, don't seem to even understand
what an incremental backup is supposed to be, and fail the basic
test of even restoring the system to the state it was in at the time
of backup.
Ugh, I guess this means I'm back to using tar, and writing my own
backup software. I have filed a bug report with SGI, so we'll see
where that goes.
Dave
--
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
