[Gluster-devel] Zombie UIDs and GIDs

[Martin Kohn]

Sorry for my late answer, the problem is solved:

Connection via Inifiniband was broken and causes the inaccessible data.
Striped volume is used because of speed and space, data loss is acceptable in our case.

Thanks for the help.
-martin

-------------------------------------------------
Martin Kohn                   
martin.kohn at cern.ch
Mobile 016094817580
-------------------------------------------------
-----Ursprüngliche Nachricht-----
Von: Vijay Bellur [mailto:vbellur at redhat.com] 
Gesendet: Freitag, 15. November 2013 06:13
An: Martin Kohn; gluster-users at gluster.org; gluster-devel at nongnu.org
Betreff: Re: [Gluster-devel] Zombie UIDs and GIDs

On 11/06/2013 11:23 PM, Martin Kohn wrote:
> Hello,
>
> I have a severe problem with gluster:
>
> User and Group IDs are broken, but not in all directories:
>
> drwxr-xr-x 103 p_kohn01 p0wess 4096 24. Jul 17:22 
> 2013_07_24_17:22_10kupsilon_flat
>
> ?---------   ? ?        ?         ?             ?
> 2013_08_10_00:44_500k_upsilon_flat
>
> ?---------   ? ?        ?         ?             ?
> 2013_08_12_21:54_500k_upsilon_flat_1s
>
> drwxr-xr-x   2 p_kohn01 p0wess    6  6. Nov 18:48 test
>
> On the Internet I could not find a hind of a solution. The volume is 
> striped over two server. In the base directorys of the bricks, the 
> data looks fine.
>
> Any Idea what I could do?
>

Is the client connected to all bricks? This seems to be the result of an incomplete readdir operation.

Is there a special need to use striping? As Lala mentioned in a different post on this thread, having a distributed or distributed - replicated volume is preferred in most deployments.

- Vijay