[Gluster-users] Again on GlusterFS and active/active WAN Replication

[Vijay Bellur]

On 02/17/2014 04:42 PM, Gionatan Danti wrote:
> On 02/17/2014 11:18 AM, Vijay Bellur wrote:
>>
>> write-behind can help with write operations but the lookup preceding the
>> write is sent out to all bricks today and hence that affects overall
>> performance.
>>
> Ok, this is in-line with my tests and the auto-generated configuration
> files
>>
>> Not as of today.
>>
>> One possibility is to let local clients assume that the remote target is
>> not reachable and hence all operations happen locally. self-heal-daemon
>> can perform background syncs when they notice that there is a delta.
>> Achieving this form of near synchronous replication will involve some
>> amount of work. If the same file is updated from multiple sites, then
>> there are chances of running into split-brains and we would need good
>> conflict resolution mechanisms too.
>>
> I thought the same thing, but what scary me is that I expect split-brain
> to be quite common, even when using frequent volume heal cycles (eg:
> each 15 min). The problem is that users are likely to work with the same
> office files. From what I understand, a split-brain scenario led to
> unaccessible files until the split-brain condition is solvede (which
> implies manual operation at the hard-link level). Is that true?

Yes, split-brains need to be resolved manually. We are exploring the 
possibility of evolving policies to resolve certain split-brains 
automatically.

-Vijay