Re: [rdiff-backup-users] Failing on 11GB file - other largefilesworking

[Ty! Boyack]

Ben Escoto wrote:
> 
> On Tue, 2 Sep 2003 11:17:00 -0700 Wolfe Maykut <address@hidden> wrote:
> > python: ERROR: (rs_file_copy_cb) seek failed: Invalid argument
> > python: ERROR: (rs_job_complete) patch job failed: IO error
> > UpdateError admin/datafax librsync error 100 while in patch cycle
> 
> Are you running librsync 0.9.6?  This problem was supposed to be fixed then.

Ben and Wolfe,

librsync 0.9.6 has mostly fixed this problem, but not quite all the
way.  The test situation I have is a user who is backing up their PC
to our network disk server (linux, rdiff-backup backs from one ext3
disk to another ext3 disk).  The user created a backup file named (for
example) backup.bkf.  This file was 12G and was working fine with
version 0.9.6 and rdiff-backup version 0.12.3.  Subsequent backups of
the same disk seemed to work fine as well.  Then the user made
substantial changes to the PC (deleted lots of files), and made a new
backup, overwriting the old backup.bkf.  The new backup.bkf file is
~7GB.  Now rdiff-backup is failing on that with:

python: ERROR: (rs_file_copy_cb) unexpected eof on fd10
python: ERROR: (rs_job_complete) patch job failed: unexpected end of
input
UpdateError backup dir/backup.bkf librsync error 103 while in patch
cycle

So it seems that so long as there are blocks in both input files (A
and B) it can create a diff file output.  But if A or B is
substantially shorter than the other, something fails.  I have not had
time to test this with more cases, so I've not brought this up before
now.  But this may be an issue that won't be solved by upgrading to
0.9.6.

Ben, if you know of an easy fix, that would be great.  Otherwise I'll
try to verify that this is consistantly the case, but I won't be able
to get to that for several days yet.

-Ty!


-- 
-===========================-
  Ty! Boyack
  NREL Unix Network Manager
  address@hidden
  (970) 491-1186
-===========================-