Wrong message when file system is full (restore)

Bug #332497 reported by Huygens
8
Affects Status Importance Assigned to Milestone
Déjà Dup
Confirmed
Low
Unassigned

Bug Description

I was attempting to restore a backup made previously on Amazon S3, just to verify that the workflow backup then restore is reliable.
I did not check before if my hard disk could accommodate the space for the backup, which it could not. But here is the error message I got, which is clearly not helpful and should be updated:
  Restore destination directory /tmp/deja-dup-5BgadS already exists.
  Will not overwrite.

That said, I have 2 other remarks about this process, but I will create 2 separate bug reports, so that they can be handled and updated separately.

Revision history for this message
Michael Terry (mterry) wrote :

Hmm, that is a bad error message! Thanks for the catch. The message should definitely be improved.

Thank you for your series of bugs about restoring. It seems Déjà Dup wasn't quite 100% reliable for you, which must not be very reassuring. :-/ The worst time for a backup solution to fail is when you are trying to restore from it.

Changed in deja-dup:
importance: Undecided → Low
status: New → Confirmed
Revision history for this message
Huygens (huygens-25) wrote :

That was fine :-) I was just testing the software to see its limit. So it was not a real case recovery. Now that I know the limits, I can plan reliably my backups as I know the "known issues" and how to avoid them. While waiting for the improvements ;-)

Do not count really on me for coding, but anyway next week-end I might have some time to have a look at the code, and if I can be of any help I will try.

Revision history for this message
Michael Terry (mterry) wrote :

That would be great. :) If you are interested, contact the ~deja-dup-hackers team mailing list or grab me on Freenode (user mterry).

Revision history for this message
Pete Goodall (pgoodall) wrote :

I'm not sure if this is the same bug or not. I'm trying to restore just one directory from my backup to test if restore will work. I am currently running the Ubuntu 9.04 Alpha 5 live CD and am logged in with the default user account (ubuntu). I installed deja-dup and ran `deja-dup --restore /home/pgoodall/bin/` from the command line. After about five minutes of eating up my cpu I receive an error that the directory already exists. There is no /home/pgoodall/bin/ directory or /home/ubuntu/bin/ directory. Also, it cannot be a space issue because according to `df -h` I have plenty of space. I deliberately chose ~/bin/ because it doesn't have much in it at all. I also tried ~/Source and ~/Documents, but I'm pretty sure the ~/Documents _is_ too big to restore in a live CD environment.

Here is what I got from the command line:

$ deja-dup --restore /home/pgoodall/bin/
** (deja-dup:10737): DEBUG: DuplicityInstance.vala:115: Running the following duplicity (10744) command: ionice -c3 duplicity collection-status file:///media/Backup/DejaDup --verbosity=9 --volsize=5 --log-fd=20

** (deja-dup:10737): DEBUG: DuplicityInstance.vala:426: duplicity (10744) exited with value 0

** (deja-dup:10737): DEBUG: DuplicityInstance.vala:115: Running the following duplicity (11122) command: ionice -c3 duplicity restore --restore-time=2009-03-02T10:32:52Z --file-to-restore=home/pgoodall/bin file:///media/Backup/DejaDup /tmp/deja-dup-OiHMQP/home/pgoodall/bin --verbosity=9 --volsize=5 --log-fd=19

** (deja-dup:10737): DEBUG: DuplicityInstance.vala:162: duplicity (11122) process killed

There seems to be plenty of room, so what is the problem?

$ df -h
Filesystem Size Used Avail Use% Mounted on
tmpfs 1007M 2.0M 1005M 1% /lib/modules/2.6.28-8-generic/volatile
tmpfs 1007M 2.0M 1005M 1% /lib/modules/2.6.28-8-generic/volatile
tmpfs 1007M 0 1007M 0% /lib/init/rw
varrun 1007M 104K 1006M 1% /var/run
varlock 1007M 0 1007M 0% /var/lock
udev 1007M 128K 1006M 1% /dev
tmpfs 1007M 92K 1006M 1% /dev/shm
rootfs 1007M 576M 431M 58% /
/dev/sr0 697M 697M 0 100% /cdrom
/dev/loop0 673M 673M 0 100% /rofs
tmpfs 1007M 12K 1007M 1% /tmp
/dev/sdb1 230G 115G 103G 53% /media/Backup

Revision history for this message
Michael Terry (mterry) wrote :

Hi, Pete!

There could be two problems. Either Deja Dup is screwing up putting the file in /tmp/deja-dup-*/home/pgoodall or duplicity is doing something odd (like running out of space).

So to help debug which and what's going on...

(A) Does the /tmp directory that Deja Dup mentions in its console output exist (any part of it)?

(B) Does the target directory (/home/pgoodall) exist and is it writable? I could imagine that the LiveCD hasn't created that and you're running as an unpriviledged user.

(B) If the above doesn't shed any light, can you post the last ~200 lines of doing: 'DEJA_DUP_DEBUG=1 deja-dup --restore /home/pgoodall/bin/ > /tmp/deja-dup.log'?

Revision history for this message
Pete Goodall (pgoodall) wrote :

Ok. Two comments:

1) When I created a user account with the same login in the live CD environment, then logged in as that user I was able to restore a reasonably sized (<1GB) directory.

2) IMO, a fundamental flaw in your logic is to restore all files to /tmp/deja-dup-${random}/. In my case I have separate '/' and '/home' partitions. As my root partition is 10 GB and my home directory is >200 GB there is no way you can cache all my files (actually only totaling ~40 GB) in /tmp.

In the second case Deja Dup aborts the restore (after quite a while of untaring as many files as it can into /tmp) with the following error:

"Restore destination directory /tmp/deja-dup-NjzY3e already exists.
Will not overwrite."

The error does not at all reflect the actuality of the problem.

I think this is a separate bug, so I will file it as such, and put the bug number in the comments here.

Revision history for this message
Pete Goodall (pgoodall) wrote :

Filed the second use case at Bug #337225

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.