In a replication cluster, one of the nodes fails. I replace it with a new machine with the same hostname and address as the failed node. Then, with glusterd stopped on the replacement and running on the survivors
- Get UUID of failed node from
/var/lib/glusterd/peers
on one of the survivors - Give it to the replacement node by putting it in
/var/lib/glusterd/glusterd.info
- Start glusterd on the replacement
- On replacement, peer probe one of the survivors
- Restart glusterd on replacement