Gluster users - Dec 2014 - Folder disappeared on volume, but exists on bricks.

Hi!

I've just had an interesting issue with our gluster storage:

I've rsync'ed data onto a mounted gluster volume, when (for some yet
unknown reason), the mountpoint of the volume got temporarily
disconnected.

Now the directory I was copying exists on the bricks, but doesn't show up
on the mounted volume.
I found a post on gluster-users@ by Franco Broi (Oct 16th) [1], but trying
to access the missing folder directly (using "ls") did nothing.


What can I do to have it accessible on the Gluster volume again?

Versions:
* GlusterFS v3.4.2
* RHEL 6.5


Thanks in advance,
Peter B.


== References:
[1] http://gluster.org/pipermail/gluster-users/2014-October/019131.html

My first approach would be taking the files from the bricks and copying
them back to the volume. Have you tried that?

2014-12-01 11:04 GMT-03:00 Peter B. <pb at das-werkstatt.com>:
> Hi!
>
> I've just had an interesting issue with our gluster storage:
>
> I've rsync'ed data onto a mounted gluster volume, when (for some
yet
> unknown reason), the mountpoint of the volume got temporarily
> disconnected.
>
> Now the directory I was copying exists on the bricks, but doesn't show
up
> on the mounted volume.
> I found a post on gluster-users@ by Franco Broi (Oct 16th) [1], but trying
> to access the missing folder directly (using "ls") did nothing.
>
>
> What can I do to have it accessible on the Gluster volume again?
>
> Versions:
> * GlusterFS v3.4.2
> * RHEL 6.5
>
>
> Thanks in advance,
> Peter B.
>
>
> == References:
> [1] http://gluster.org/pipermail/gluster-users/2014-October/019131.html
>
>
>
>
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://supercolony.gluster.org/mailman/listinfo/gluster-users
>


-- 
Pavlik Salles Juan Jos?
Blog - http://viviendolared.blogspot.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://supercolony.gluster.org/pipermail/gluster-users/attachments/20141201/fe11bce7/attachment.html>

I think I know now what happened:

- 2 gluster volumes on 2 different servers: A and B.
- A=production, B=testing
- On server "B" we will soon expand, so I read up on how to add new
nodes.
- Therefore on server "B" I ran "gluster probe A", assuming
that "probe"
was just reading *if* A would be available.

It seems that "probing" another node already writes configuration data
onto them.


That seems to be the reason why the data of the bricks didn't show up on
the volume mountpoint on "A", because it was now already pointing to
"B".


Pb