On Tue, Dec 14, 2004 at 10:35:00AM -0600, Aleksandar Milivojevic wrote: > James Wilkinson wrote: > >It's worth noting that Linus Torvalds believes the program to be > >terminally broken: > >http://lwn.net/2001/0503/a/lt-dump.php3 > > > >He writes: > > So anybody who depends on "dump" getting backups right is already > > playing russian rulette with their backups. > > Of course he wrote that. He probably felt some guilt for not providing > both block and character devices for disks, as any other half-decent > Unix/Unix-like OS has ;-) All that aside a live file system when viewed from the bits on the disk is not a complete view of things. There is stuff in memory, in the journal and on disk. A backup tool that can only see the bits on the disk will make errors because of the missing information and dynamic nature of the flawed view. Having said that 'dump' works because most information and data on the disk is not dynamic. Also most of the structures terminate and will not throw dump in infinite loops. One potentially serious risk of dump on a live files system is data reuse policy. Some folks care about this more than the data errors that might be expected. It should be possible to modify tar and cpio to archive security context acls, attributes and other extended meta data. (NTFS needs to be included in discussions on this topic too.) In addition tools like tripwire should be security context, files system, ACL and attribute aware. -- T o m M i t c h e l l spam unwanted email. SPAM, good eats, and a trademark of Hormel Foods.
