add committed index in Progress #366
Conversation
Signed-off-by: accelsao <[email protected]>
|
I think my test_case is useless.. |
Signed-off-by: accelsao <[email protected]>
Signed-off-by: accelsao <[email protected]>
|
error |
Signed-off-by: accelsao <[email protected]>
src/progress/mod.rs
Outdated
| @@ -119,6 +123,7 @@ impl Progress { | |||
| self.recent_active = false; | |||
| debug_assert!(self.ins.cap() != 0); | |||
| self.ins.reset(); | |||
| self.committed_index = 0 | |||
| @@ -212,6 +212,121 @@ fn test_progress_update() { | |||
| } | |||
| } | |||
|
|
|||
| #[test] | |||
| fn test_progress_committed() { | |||
There was a problem hiding this comment.
I think you can use Network to test the commit index change. It should validate following case:
- Log rejection should not change commit index
- Successfully Append should update commit index
- Heartbeat should update commit index
- Commit index won't step backward.
There was a problem hiding this comment.
I still do not know when will we get the smaller commit index
There was a problem hiding this comment.
For example, leader sends two MsgAppend to follower and gets MsgAppendResponse in reverse order.
Signed-off-by: accelsao <[email protected]>
Signed-off-by: accelsao <[email protected]>
Signed-off-by: accelsao <[email protected]>
Signed-off-by: accelsao <[email protected]>
Signed-off-by: accelsao <[email protected]>
|
|
||
| nt.send(vec![msg_append_response.clone()]); | ||
|
|
||
| assert_eq!(nt.peers[&1].prs().get(2).unwrap().committed_index, 10); |
There was a problem hiding this comment.
shouldnt that commiitted index is update as 10?
There was a problem hiding this comment.
If leader doesn't have log 10, it shouldn't commit to 10. Giving that raft-rs doesn't take much effort in byzantine fault tolerant, I think remove the case is OK for now.
Signed-off-by: accelsao <[email protected]>
Signed-off-by: accelsao <[email protected]>
Signed-off-by: accelsao <[email protected]>
Close #325
Signed-off-by: accelsao [email protected]