[Gluster-devel] Glusterd: A New Hope

[Daniel Mons]

As an end user and non-developer, I don't have much to add from the
dev side of things.  However all of the discussion in this thread has
prompted me to throw in my 2c.

As a sysadmin running 856TB of GlusterFS in production, I believe the
3.X way of dealing with things (glusterd, dynamic) is far
easier/better than the 2.X way (config files, static).  We have a
comprehensive Puppet setup in place, so configuration management isn't
the issue for us (we solved that problem ages ago).  The problem is
being able to make dynamic changes to a running system without
affecting users working on the system.  Our business is 24x7 (like so
many these days), and downtime (scheduled or otherwise) needs to be
kept to a minimum just to survive the current economic climate.

I flat out do not agree with the "3.X is bloatware, go back to 2.X
config files" points above.  Indeed, the change to a more dynamic
3.X/glusterd model was exactly what prompted me to take GlusterFS
seriously as a real world solution to our storage problems.  Even
within the first few weeks of a production rollout, the dynamic
management aspects of 3.X saved us a world of pain, and kept our
business moving.

That's all I really have to add.  I'm watching this thread with
interest.  It sounds like Gluster dev is in good hands, and the right
sorts of discussions are being had out in the open.  My confidence in
the project remains very high.

-Dan