Gluster users - Jul 2015 - Replacing a replicated brick

On 07/27/2015 03:38 PM, Iain Milne wrote:
> Hi,
> 
> Hopefully a fairly simple question:
> 
> If I have a 3.6 distributed replica-2 setup with four nodes of one brick
> each (so A/B in replicate as are C/D) what is the process of commands I
> need to run to replace B, with new brick E?
You can go for replace brick commit force option. Please note commit
force can result into data loss. You would need to ensure that you copy
the data from the failed brick to the new brick
manually.
> 
> It'd be handy to know the procedure for the cases when B was both on
and
> offline.
> 
> Thanks
> 
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://www.gluster.org/mailman/listinfo/gluster-users
> 
-- 
~Atin

On Monday 27 July 2015 04:53 PM, Atin Mukherjee wrote:
>
>
> On 07/27/2015 03:38 PM, Iain Milne wrote:
>> Hi,
>>
>> Hopefully a fairly simple question:
>>
>> If I have a 3.6 distributed replica-2 setup with four nodes of one
brick
>> each (so A/B in replicate as are C/D) what is the process of commands I
>> need to run to replace B, with new brick E?
> You can go for replace brick commit force option. Please note commit
> force can result into data loss. You would need to ensure that you copy
> the data from the failed brick to the new brick manually.
>>
If you ensure that the brick being replaced is not the source for any 
pending self-heals, replace-brick commit force in a replicated volume is 
safe. self-healing would take care of re-populating data once a new 
brick is in place.

Regards,
Vijay