[Gluster-users] [ovirt-users] gluster split-brain on vm images volume

[Pranith Kumar Karampuri]

On 08/03/2014 01:43 AM, Tiemen Ruiten wrote:
> On 08/02/14 20:12, Pranith Kumar Karampuri wrote:
>> On 08/02/2014 06:50 PM, Tiemen Ruiten wrote:
>>> Hello,
>>>
>>> I'm cross-posting this from ovirt-users:
>>>
>>> I have an oVirt environment backed by a two-node Gluster-cluster.
>>> Yesterday I decided to upgrade to from GlusterFS 3.5.1 to 3.5.2, but
>>> that caused the gluster daemon to stop and now I have several lines like
>>> this in my log for the volume that hosts the VM images, called vmimage:
>> Did the upgrade happen when the volume is still running?
> Yes...
I guess we need to document upgrade process if not already done.
>>> [2014-08-02 12:56:20.994767] E
>>> [afr-self-heal-common.c:233:afr_sh_print_split_brain_log]
>>> 0-vmimage-replicate-0: Unable to self-heal contents of
>>> 'f09c211d-eb49-4715-8031-85a5a8f39f18' (possible split-brain). Please
>>> delete the file from all but the preferred subvolume.- Pending matrix: [
>>> [ 0 408 ] [ 180 0 ] ]
>> This is the document that talks about how to resolve split-brains in
>> gluster.
>> https://github.com/gluster/glusterfs/blob/master/doc/split-brain.md
> OK, I will try that.
>>> What I would like to do is the following, since I'm not 100% happy
>>> anyway with how the volume is setup:
>>>
>>> - Stop VDSM on the oVirt hosts / unmount the volume
>>> - Stop the current vmimage volume and rename it
>> Is this a gluster volume? gluster volumes can't be renamed..
> That surprise me: in the man page I find this:
>
> volume rename <VOLNAME> <NEW-VOLNAME>
>                Rename the specified volume.
One more documentation bug :-(
>
>>> - Create a new vmimage volume
>>> - Copy the images from one of the nodes
>> where will these images be copied to? on to the gluster mount? if yes
>> then there is no need to sync.
> OK, I will try to resolve with the guide for split-brain scenarios first.
Let us know if you have any doubts in this document.
>>> - Start the volume and let it sync
>>> - Restart VDSM / mount the volume
>>>
>>> Is this going to work? Or is there critical metadata that will not be
>>> transferred with these steps?
>>>
>>> Tiemen
>>>
>>>