Backups - simple question

Dave Olson olson at anchor.esd.sgi.com
Wed Jun 5 06:15:02 AEST 1991


In <9106032216.AA22384 at koko.pdi.com> shoshana at pdi.UUCP (Shoshana Abrass) writes:


|   Once or month or so, we do an 'archive' backup to cartridge tapes 
|   (as opposed to exabyte). The backup takes about 5 cartridges.
|   If one of the cartridges has a bad spot, it does what I think is 
|   the wrong thing.  For example, tape 1 and 2 have been successfully
|   written. Files are being written to tape 3. Suddenly the driver
|   encounters a write error and asks for a new tape. If I put in a new
|   tape, it starts writing files where it left off from tape 3. What I 
|   want it to do is rewrite all of tape 3 on the new tape. I have no 
|   interest in having a tape with bad media in one of my archive sets, 
|   plus it screws up offsite storage if I can't predict how many tapes 
|   my backup will end up being. 
| 
|   Does this bother anyone besides me? or maybe I'm not doing the right 
|   things with bru?  I use the -s size option, which doesn't seem to
|   help. Using dump is not a possibility. Our tapes are data quality 
|   tapes from inmac. 

Well, presumably if the media was good up to that point, then it
is still good.  The backup (and restore!) should work correctly.
If you don't like this behavior, about all you can do is pretest
your media, or abort the backup and restart.

None of the shipped backup programs at SGI keep track of where the
volume started (except on the first i/o), so there is no way to
restart the volume.  There may be commercial packages that do
what you want, but I don't know of any.  Think of the problems
of keeping all the state around about where in the tree (and
potentially in a file) you were at the start of the volume.
Since most of the backup programs use the stack to keep track
of where they are in the tree (in order to deal with symlinks)
this isn't a trivial project, although certainly not impossible.
--

	Dave Olson

Life would be so much easier if we could just look at the source code.



More information about the Comp.sys.sgi mailing list