[Gluster-users] Convert From Distributed-Replicate to Distributed-Only Then Back Again

[Ken Schweigert]

I have a 4-node distributed-replicate setup running 3.5.2 that I am trying
to get synced up with an existing NFS solution; I want Gluster to replace
NFS once they are both in sync.  The problem that I have is that data is
written to the NFS server faster than rsync can copy to Gluster so I'm at a
point to where I'm always about a day behind.  I think if I can get this
intermediate step of syncing out of the loop, and just have our web
application write straight to gluster, then I think I can get caught up and
get rolled in to this storage solution.

I think that maybe the extra writes to the replication pair might be
contributing to this initial problem.  I would like to find out how I could
convert it to just be a distributed configuration, get everything caught up
and in sync, then bring the replication pair back online and go back to the
original setup.

If I understand it correctly, the self-heal process is what takes care of
the replication piece and is something that always runs.  If this is
correct, shouldn't I be able to just power down the pair of servers that
are the replication pair, let the sync get caught up by writing to the
distributed pair, then restart those servers and let self-heal bring the
replication pair back online?

Does anyone see anything wrong with doing it this way or is there a better
way to do something like this?  Are there any potential "gotchas" that I
should be aware of?

-ken
-- 
Have a nice day ... unless you've made other plans.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20150415/9eeaf566/attachment.html>