Move replication algorithm to Engine #650
Comments
|
👋 Thanks for opening this issue! Get help or engage by:
|
drmingdrmer
added a commit
to drmingdrmer/openraft
that referenced
this issue
Jan 22, 2023
Let Engine totally decide what to do, ReplicationCore becomes a passive executor. Engine emit a replication request and let ReplicationCore execute it. But it is not an exact one request one reply model, e.g., Engine does not expect the reply for a broadcast-committed request. Engine keeps the track of every inflight replication request, and update the `inflight` state when a reply is received from ReplicationCore. A new type `Inflight` is introduced to manage the state of inflight replication data, e.g., a series of log entries or a snapshot. ReplicationCore does not need to ask for RaftCore for snapshot anymore. But instead, if snapshot replication is needed, Engine will just send a snapshot to ReplicationCore. Introduce `ReplicationHandler` and `LogHandler` as a sub set function container of Engine, in order to make `Engine` not that big. Add more `Validate` implementations, to assert internal consistency. Add ProgressEntry::next_send() to calculate what entries/snapshot to send Remove ProgressEntry.mid; It can be calculated when needed. - Fix: datafuselabs#650
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
No description provided.
The text was updated successfully, but these errors were encountered: