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

[Christopher Pereira]

On 11-05-2015 17:57, Vijay Bellur wrote:
> On 05/11/2015 10:45 PM, 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".
>
> Zero byte files should be present inside the 3rd brick for maintaining 
> afr's metadata. Do you happen to see files of similar size & consuming 
> same disk space in the 3rd brick as in the first and second bricks?

Hi Vijay:

Thanks for answering.

"du" reveals the 3rd brick has the same size as the others (over 1 GB), 
so something is wrong.

>
>
>> Also, I don't see the .meta dir inside the mount point.
>
> You will not be able to see .meta but should be able to change 
> directory to .meta from the mount point. Please let us know if that is 
> not the case.

Got it.
I confirm that .../options/arbiter-count contains "1".