Hm, my experience with OpenEBS has been... well, not that good. I set up a small cluster and tried a catastrophe scenario by just removing a node. Then i would

Hm, my experience with OpenEBS has been... well, not that good. I set up a small cluster and tried a catastrophe scenario by just removing a node. Then i would either get processes hanging in D state, or blazingly fast i/o, which, upon bringing up the node back would result in data loss. Keeping in mind that it was just a simple bash script counting from 1 to infinity to / from a file and OpenEBS couldn't cope with that, i opted for rook and haven't looked back since. It actually runs under /var/lib/rook by default and losing a node or two has no impact, provided your replica count is sufficient. Rook really manages the underlying ceph cluster good most of the time (until you start shuffling osds around), and ceph itself is a wonderful piece of software - it really shows that it came from CERN. But to each his own, i guess.