Backup & Misc

[Richard Todd]

jim at jagmac2.gsfc.nasa.gov (Jim Jagielski) writes:

>I seem to recall somewhere in the documentation for 1.1, that one of the
>considerations when using dump and restore was that the restored backup
>had to be the same size file system that was originally dumped. In other
>words, if your file system was orginally 40 megs and you wanted to "increase"
>it to 80 megs, the documentation suggested NOT using dump/restore since the
>file system sizes were different and it would not work. (they also mentioned
>in there information about the SB being copied also)

  Well, the documentation was wrong, then.  I don't have an A/UX 1.1
manual set, but I do have a set of A/UX 1.0 manuals.  (The 1.0 manuals
did document dump/restore, even though those programs apparently
weren't finished in time for the 1.0 release!)  Anyway, the 1.0 man
page for restore says:
     "A dumpfs(1M) followed by a mkfs and a restore is used to change the size
of a file system." 
  For some reason, this helpful statement isn't in the online A/UX 2.0 manuals.
It obviously should be.  Sigh.  
	
>Now maybe the documentation was wrong, or maybe dump/restore was changed/fixed
>to circumvent this (heck, I remember the days of nightly "dd" backups!) under
>2.0. I simply felt the need to relay what I read, even if it IS/WAS wrong :)

  Quite understandable--it isn't your fault your manual lied to you...

  My understanding is that dump/restore have worked this way ever since, at 
the least, some early BSD release.  The deal is that while dump has to
and does grovel through the raw disk, restore doesn't do so--all the restoring
of files is done through normal file I/O calls.  Thus restore doesn't control
the specific allocation of a file to a specific inode or block number, so
restore doesn't really care if the filesystem it's restoring onto has different
parameters than the one originally dumped.  (I recall hearing that V7 restor 
(note the spelling) *did* write to the raw fs, but I don't know for sure; in
any case, I don't think any version of BSD restore did.) 
--
Richard Todd	rmtodd at uokmax.ecn.uoknor.edu  rmtodd at chinet.chi.il.us
	rmtodd at servalan.uucp
"Cancelling a posted message means posting a cancel message."-Maarten Litmaath