Bug: Follower not starting election timeout

[lionesswardrobe]

Seems to happen when follower was previously transitioned from the leader state.
I'm not too familiar with the code but briefly looking at here:
https://github.com/railgun-rs/actix-raft/blob/fd0e34c2e133ea9950aa463ba4c30a824d2055be/src/raft/mod.rs#L291-L296

Should
if let Some(handle) = self.election_timeout {
here actually be
if let Some(handle) = self.election_timeout.take() {
instead?

[thedodd]

Hey @lionesswardrobe thanks for the bug report. Looking at that code, it looks like you are correct. I believe this is a compound issue where for some given node A.

• node A was a follower.
• node A was elected as leader (and due to the code above does not properly remove the election timeout state).
• node A becomes a follower again and then fails to start the election timeout due to the above failure to use .take(), which combined with this line in the become_follower handler:
  https://github.com/railgun-rs/actix-raft/blob/fd0e34c2e133ea9950aa463ba4c30a824d2055be/src/raft/mod.rs#L206-L209
  the follower fails to start a new election timeout loop. Solid bug report!

Here is my plan of action:

• update all of the locations in the code where this pattern is repeated (take the election_timeout handle and cancel it) and encapsulate it in a method so that it is not missed.
• write a few tests to ensure this behavior.
• cut a new release. Should be 0.4.3. There are already a few staged changes which will go out as part of the release as well. This is the only bug fix as part of that release though :).

Thanks again @lionesswardrobe!

[thedodd]

Cutting 0.4.3 now.

[thedodd]

@lionesswardrobe just wanted to circle back with you here. Did 0.4.3 resolve the issue you were seeing?