Hello,
I am trying to understand how membership change works in etcd-raft and have some questions here.
Correct me if I am wrong, it seems etcd-raft does not explicitly persist the raft topology, but instead treat each membership change as a raft log entry. I am looking at the node implementation of etcd-raft. Apparently, when starting a new node, it
manually injects a configChangeEntry; and when restarting a node, it relies on replaying WAL logs from the very beginning to obtain peer information.
However, what happens if we have a persisted appliedIndex, in which case raft won't re-apply all the logs from index 1? How does it know its peer information in order to join the group?
Thanks a lot.
Jia