Closed pihme closed 4 years ago
Looks like we are testing a behavior that does not exist :smile: Our snapshot replication is not reliable - that is - if a node is not reachable during the snapshot replication it won't receive it. There are no retries. But this test is testing that the node will receive it when it comes back. Test sometimes passes because restart is fast and the broker is up before the other node replicates snapshot.
We can remove this test. There is already a test for testing restarted node receives snapshot when there is a new snapshot.
I think the goal here was to test the Raft replication. Can we restart the follower with data loss so we do send it the latest snapshot? Or do we have tests for this?
I think the goal here was to test the Raft replication. Can we restart the follower with data loss so we do send it the latest snapshot? Or do we have tests for this?
Aren't we testing in RestoreTest?
You're right, thanks! :)
Summary
Failures
Example assertion failure
Hypotheses
Logs
Logs