Re: [rdiff-backup-users] Warning: Extended attributes file not found

[Steve White]

Andrew,

On  8.01.08, Andrew Ferguson wrote:
> Steve White wrote:
> >> The Warning message indicates that rdiff-backup cannot find the Extended
> >> Attributes storage file, which is normally kept in rdiff-backup-data/ .
> >> So yes, it is saying which file triggered the problem.
> >> 
> > I don't follow.  As I said, no file name was reported.
> > 
> > In what way is it saying which file triggered the problem?
> > 
> > Often I get a report just like this:
> > 
> > Warning: Extended attributes file not found
> 
> The message is:  "Extended attributes" file not found.
> 
> The file in which rdiff-backup stores the extended attributes could not
> be found. This is an unusual error, because it means that rdiff-backup
> didn't store extended attributes in the past, but is now looking for them.
> 
> Make a small test backup on your client. (eg, "rdiff-backup test
> test-backup"). Then look in the "test-backup/rdiff-backup-data/" folder
> -- you will probably see a file
> "test-backup/rdiff-backup-data/extended_attributes.2008-....." --- that
> is the "Extended attributes" file.
> 
Ah.

I did your test, on both the client and server.  On both I see such 
"extended_attributes*" files.

I have a recommendation:  the message ought to read
        Warning: "extended_attributes*" file not found.

> >> Did you perhaps change your rdiff-backup configuration between runs? 
> > 
> > I don't think the rdiff-backup configuration changed.
> 
> Then something else has changed, such as the pyxattr module has been
> installed, or extended attributes have been enabled on a volume.
> 
My best guess is that something changed on the target machine, where I 
recently upgraded the operating system there.  But I didn't change
the filesystems in question.

> >> Do you have extended attributes setup on the server but not the client?
> > 
> > How would I know that?
> 
> 
> Well, what OS are the server and client running on? On Mac OS X, you
> have to enable them by hand on 10.4 client, but they are enabled by
> default on 10.4 server; not sure about 10.5. On linux, you use the
> user_xattr option when mounting ext3. Not sure about other options.
> 
Would that be in the fstab file?  I don't see anything like that. 
It just uses the "defaults" for ext3.   According to "man mount", that
is just
        rw, suid, dev, exec, auto, nouser, and async

But I really suspect it was the recent upgrade that did this (from 
Scientific Linux 4.5 to 5.0).  There is often a big mess in the 
upgrade involving partition labels.


> See what 'mount' tells you on both the client and server. However, all
> of this information will be reported by "rdiff-backup -v5 ..."
> 
Well I tried
        mount -l
but it didn't tell me much.  The partitions are all ext3, all mounted
essentially the same, for instance:

Target:
        /dev/sdb6 on /work3 type ext3 (rw) [/work3]
Source:
        /dev/sdb2 on /work3 type ext3 (rw) []

Well... could the partition label be evidence of extended attributes?

Cheers!

-  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  - 
Steve White                                             +49(331)7499-202
Computer Programmierer                                    Zi. 35  Bg. 20
-  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  - 
Astrophysikalisches Institut Potsdam (AIP)
An der Sternwarte 16, D-14482 Potsdam

Vorstand: Prof. Dr. Matthias Steinmetz, Peter A. Stolz

Stiftung privaten Rechts, Stiftungsverzeichnis Brandenburg: III/7-71-026
-  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -  -