[Gluster-devel] Avoid split-brains on replica-2

[Ravishankar N]

On 05/12/2015 02:15 AM, Christopher Pereira wrote:
> On 11-05-2015 15:40, Christopher Pereira wrote:
>>
>>> There is an arbiter feature for replica 3 volumes 
>>> (https://github.com/gluster/glusterfs/blob/master/doc/features/afr-arbiter-volumes.md) 
>>> being released in glusterfs 3.7 which would prevent files from going 
>>> into split-brains, you could try that out. If the writes can cause a 
>>> split-brain, it is failed with an ENOTCONN to the application.
>
> Seems to be broken, because files are also stored on the 3rd brick, 
> even when using "replica 3 arbiter 1".
> Also, I don't see the .meta dir inside the mount point.
>
> Bricks 1 and 2 are using:
>
> glusterfs-3.7.0beta1-0.131.gitf54b232.el7.centos.x86_64
> glusterfs-rdma-3.7.0beta1-0.131.gitf54b232.el7.centos.x86_64
> glusterfs-cli-3.7.0beta1-0.131.gitf54b232.el7.centos.x86_64
> glusterfs-api-3.7.0beta1-0.131.gitf54b232.el7.centos.x86_64
> glusterfs-server-3.7.0beta1-0.131.gitf54b232.el7.centos.x86_64
> glusterfs-geo-replication-3.7.0beta1-0.131.gitf54b232.el7.centos.x86_64
> glusterfs-libs-3.7.0beta1-0.131.gitf54b232.el7.centos.x86_64
> glusterfs-fuse-3.7.0beta1-0.131.gitf54b232.el7.centos.x86_64
> vdsm-gluster-4.17.0-778.git597bb40.el7.noarch
> glusterfs-debuginfo-3.7.0beta1-0.155.git72f80ae.el7.centos.x86_64
>
> Brick 3 is using:
>
> glusterfs-fuse-3.8dev-0.58.gitf692757.el7.centos.x86_64
> glusterfs-geo-replication-3.8dev-0.58.gitf692757.el7.centos.x86_64
> glusterfs-cli-3.8dev-0.58.gitf692757.el7.centos.x86_64
> glusterfs-api-3.8dev-0.58.gitf692757.el7.centos.x86_64
> glusterfs-server-3.8dev-0.58.gitf692757.el7.centos.x86_64
> glusterfs-rdma-3.8dev-0.58.gitf692757.el7.centos.x86_64
> vdsm-gluster-4.17.0-743.gite5856da.el7.noarch
> glusterfs-libs-3.8dev-0.58.gitf692757.el7.centos.x86_64
> glusterfs-3.8dev-0.58.gitf692757.el7.centos.x86_64
> glusterfs-debuginfo-3.8dev-0.58.gitf692757.el7.centos.x86_64
>
>
Hi Christopher,
The patches for arbiter logic have gone in for 3.7.0beta2. Could you 
upgrade your bricks 1 and 2 to that version or to 3.8dev like brick3 and 
try? It is recommended to have all 3 nodes on the same version. After 
creating the volume, you can check the volfile of the 3rd brick- it must 
contain the arbiter xlator in the graph. i.e. 
/var/lib/glusterd/vols/<volname>/<volname.IP.the_third_brick.vol must 
have this entry:

/volume <volname>-arbiter//
//    type features/arbiter//
//    subvolumes <volname>-posix//
//end-volume/

Thanks,
Ravi

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.gluster.org/pipermail/gluster-devel/attachments/20150512/8d6f66b9/attachment.html>