We've been using glusterFS in production for several years. It is slow and can be slow, especially when nodes experience higher network latency. When writes are synchronously executed across all the nodes we've seen higher load averages on servers.

However it has been working reliably. So for non-critical time constrained loads it's good enough.