[Gluster-users] going from replca 2 to replica 3 for a while? Or replace bricks?

[Matthew Nicholson]

So, I've got a question that is more "best practice" based or I suppose
"what would you do?":

Currently:

2 Datacenters, A and B, with 10x2 distributed replicated gluster volume.

A01 <=> B01
A02 <=> B02
...
A10 <=> B10


We are bringing a third datacenter, C, online and B wil be going away. We
are getting another 10 bricks for datacenter C, and want to, while keeping
the storage available, basically move B to C. So, the end state would be:

A01 <=> C01
...
A10 <=> C10

(and then we would take the BXX bricks and add half to A and half to
C, but thats just simple add bricks one things are re-balanced) .

So, I see a couple ways to do this:

1: if possible, add C01-C10 to the volume, as a third replica, and then
once the heal is complete, remove all the B bricks and reduce teh replica
count back to 2...

2: do brick-replace on each BXX brick with the new CXX brick, for each
brick. replica count stays at 2, and we just move the replica to C one
brick at a time.

So, with option 1, i guess the question is...is it even possible to add 10
more bricks while upping the replica count? Likewise, to remove 10 and
decrease the replica count?

Option 2, as far as I know should work, but I'm trying to gauge other
options over ding brick by brick replacements for this much data.

Anyone out there done something sort of like this?




--
Matthew Nicholson
Research Computing Specialist
Harvard FAS Research Computing
matthew_nicholson at harvard.edu
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://supercolony.gluster.org/pipermail/gluster-users/attachments/20130410/52fa167c/attachment.html>