Re: lost inode, no backup

From: Randy Bush <randy_at_psg.com>
Date: Mon, 02 Jan 2012 09:17:42 +0900
[ i fear that, at this point, this only serves as a oddity i have seen
  and reported just in case someone else sees something similar and the
  two clues can be summed ]

> I know it's been a week but no-one else has offered any input.

no blame, was very very strange.  never seen anything like it before.
and i pieced my world back together, with very little loss, as there
were other modes of backup in place, belt and braces.  i only actually
lost some months of mail in imap folders of procmail mailing list spew,
such as freebsd-current.

> It's not explicitly specified but I presume /usr/home is a directory
> within /usr.  (Not a symlink or mountpoint).

indeed.

> What version were you running before 10-current?

9-[then-]current.  the machine is the one server which only has my
personal crud, so it is the bleeding edge.  it converted to -10 about
when the problems with backup content seem to have started last
(northern hemispheric) summer.  so the issue could have been some file
state that happened long ago and just did not surface until the much
later upgrade.

> Is /usr UFS2 with softupdates+journalling?

/dev/da0s1f on /usr (ufs, local, soft-updates)

> Is userland (specifically /sbin/dump and fsck) aligned with the
> kernel?

should have been.  but lots of things in life should have been.

> Any idea what SVN revision your system corresponds to?

i guess i could restore a dump and find out.

> Have there been any unclean shutdowns of the system?

quite likely.  a neighbor on the same vmware host did some exciting
resource demand messes the other month.

> "Tape blocks" reported by dump(8) are (as far as I can tell) always
> 1KB.  Your df(1) output reports 111GB used in /usr but dump(8) only
> shows ~57GB being backed up.  Do you have any idea why there's such a
> difference?

/usr/home not being backed up.  among other things, /usr/home/randy/rair
contains a unison rsync of my laptop's ~randy, i.e. it's large.

> How much space would you expect to be used in /usr?

/dev/da0s1f    128G     55G     63G    47%    /usr

> Do you have the fsck(8) output?  If so, how much space does it report?

not sure what it is you expect

# fsck da0s1f
** /dev/da0s1f (NO WRITE)
** Last Mounted on /usr
** Phase 1 - Check Blocks and Sizes
** Phase 2 - Check Pathnames
... lots'o whining
613062 files, 28926263 used, 38606650 free (134698 frags, 4808994 blocks, 0.2% fragmentation)

# du -sh /usr/home
 48G    /usr/home

> Have you checked the dumps for all the other filesystems?  Are any of
> them missing data or containing unexpectedly empty directories?

checked (very pseudo) random directories and filesystems.  all good.

> I don't suppose you have the output from hd(1), stat(1) and "ls -al"
> of /usr & /usr/home?

not the ones which were borked, only current seemingly good ones.

randy
Received on Sun Jan 01 2012 - 23:17:45 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:40:22 UTC