[ANNC] Riak 1.3.0 RC1

Sean Cribbs sean at basho.com
Thu Jan 31 09:44:21 EST 2013


Simon,

Correct, that behavior is unchanged. Note that the hash stored in the
AAE trees is a hash of the *entire* object, including metadata, vector
clock and any siblings. If two replicas diverge because of concurrent
writes, the AAE-triggered read-repair will cause the siblings to be
stored on both replicas. However, once the read-repair is successful,
they will have the same hash and not repair again.

On Thu, Jan 31, 2013 at 1:41 AM, Simon Effenberg
<seffenberg at team.mobile.de> wrote:
> On Wed, 30 Jan 2013 21:22:33 -0800
> Joseph Blomstedt <joe at basho.com> wrote:
>
>> The main issue is that Search data
>> doesn't include anything equivalent to vector clocks. If two replicas
>> have divergent data, who wins?
>
> so what happens if I have (because of a temporary split brain and a
> write to the key on both brain's) sibblings.. is the AAE only
> triggering an 'read repair' which causes the sibblings which then has
> to be repaired/merged by the next client which is reading the key?
>
> Cheers,
> Simon



-- 
Sean Cribbs <sean at basho.com>
Software Engineer
Basho Technologies, Inc.
http://basho.com/




More information about the riak-users mailing list