[net.unix-wizards] excessive resync errors from 4.2 restore

RSanders@USGS2-MULTICS.ARPA (04/12/85)

  Our monthly dumps of the user filesystem on an Eagle (~350 Mb) are
running into 8 or 9 tapes.  Recently, we've been trying to restore(8)
individual files from these dumps.  We are getting lots of 'resynch'
error messages and are unable to restore several files.

  We usually have PM on the tape drive (Cipher 920/Emulex TC-11
controller) a few days before the monthly dumps, and I just had the
maintenance people go over the tape drive with a fine tooth comb.  The
only reference in my archives to this problem points at flaky tape
drives.  I found nothing relevant in the Mt Xinu buglist handed out in
Dallas.  Has anyone got some clues?

-- Rex
   RSanders@Denver  (milnet/arpanet)

henry@utzoo.UUCP (Henry Spencer) (04/15/85)

>   Our monthly dumps of the user filesystem on an Eagle (~350 Mb) are
> running into 8 or 9 tapes.  Recently, we've been trying to restore(8)
> individual files from these dumps.  We are getting lots of 'resynch'
> error messages and are unable to restore several files.

Do you run your dumps while the system is active multi-user?  My
understanding is that 4.2 dump is better about this than the old Unix
dump, but it's probably less than perfect.  It is extremely hard to
build a dump program that will run perfectly when the file system
is changing underfoot, and I don't believe Berkeley tried.  Truly
reliable dumps require a quiescent file system, which in practice
means you'd better unmount it to make sure it's really idle.  This
generally means that you have to run your dumps single-user.
-- 
				Henry Spencer @ U of Toronto Zoology
				{allegra,ihnp4,linus,decvax}!utzoo!henry

RSanders@USGS2-MULTICS.ARPA (04/19/85)

More details on our problem -

re:  Dave Johnson's message -
  We have a 4k/512 user filesystem, and the resynch errors are real
enough to cause restore NOT to restore certain files asked for
individually from our level 0 dumps.  I don't have the capacity to
attempt a complete filesystem restore.

re:  Henry Spencer's message -
  I haven't done any kind of dump in multi-user mode since my PDP 11/34
- V7 days - even our daily incrementals are done single-user.

Other details -
  We haven't seen this problem on our daily/weekly tapes - yet.  We have
a relatively low level of requests for file restores from backup.

  We've had no complaints from tar, dd, or vmstape, from reading/writing
tapes from/to various foreign systems (as well as our own).  Some of
these systems are very foreign - tar tapes to/from Multics, for example.

-- Rex