Gluster users - Nov 2013 - Problems recovering volume

Hello everyone;

First I would like to congratulate all the gluster community for such a
project. I have been using gluster for a while and I am very happy with it
:-)

I am facing the following situation in a gluster distributed volume deploy
that currently uses gluster 3.4.0 and native gluster clients.

For some reason, which I don't know yet, configuration files
(/var/lib/glusterd/) have been damaged so the volumes config is lost but
the data remains intact.

I have been advised in #gluster that I can recreate the volumes without
losing the data. I have followed this procedure:

0.Create a backup of each brick.
1.Umounting the partitions that contain the gluster volumes.
2.Create the distributed volume again in the same mount point that it was
previously.
3.Mount the data.
4.Start the volume.

Step 4 was outputing an error regarding the volume-id so I followed the
instructions from Joe Julian's website:
http://joejulian.name/blog/glusterfs-path-or-a-prefix-of-it-is-already-part-of-a-volume/

After a while I managed to start the volume and mount the data in my
clients but I have some inconsistencies like some files are missing in some
of the bricks.

Is there any way to make the volumes consistent again/check consistency or
do I have to restore missing data?

Regards.



-- 
Diego Lendoiro
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://supercolony.gluster.org/pipermail/gluster-users/attachments/20131116/51bfe83a/attachment.html>

On 11/16/2013 04:24 PM, Diego Lendoiro wrote:
> Hello everyone;
>
> First I would like to congratulate all the gluster community for such a
> project. I have been using gluster for a while and I am very happy with
> it :-)
>
> I am facing the following situation in a gluster distributed volume
> deploy that currently uses gluster 3.4.0 and native gluster clients.
>
> For some reason, which I don't know yet, configuration files
> (/var/lib/glusterd/) have been damaged so the volumes config is lost but
> the data remains intact.
You could have possibly re-generated the configuration files by 
performing a volume set option:

#gluster volume set <volname> network.ping-timeout 42

would have created the configuration files back.
>
> I have been advised in #gluster that I can recreate the volumes without
> losing the data. I have followed this procedure:
>
> 0.Create a backup of each brick.
> 1.Umounting the partitions that contain the gluster volumes.
> 2.Create the distributed volume again in the same mount point that it
> was previously.
> 3.Mount the data.
> 4.Start the volume.
>
> Step 4 was outputing an error regarding the volume-id so I followed the
> instructions from Joe Julian's website:
>
http://joejulian.name/blog/glusterfs-path-or-a-prefix-of-it-is-already-part-of-a-volume/
>
> After a while I managed to start the volume and mount the data in my
> clients but I have some inconsistencies like some files are missing in
> some of the bricks.
Are files missing in the bricks or are they missing when seen from 
native gluster clients? If files are not being seen from the native 
clients, it is more likely that a glusterfsd server is not online. Does 
gluster volume status report that all processes are online for this volume?

Regards,
Vijay