Qbft testing #128
Qbft testing #128
Conversation
|
Thought again about the queue we are passing. I am not sure if |
|
@dknopik I was actually just planning on scrapping I think |
Well,
Alright! |
anchor/qbft_manager/src/lib.rs
Outdated
| // Queue is full - drop message under constrained bandwidth | ||
| warn!("Dropping QBFT message due to full queue: msg={:?}", msg); | ||
| } | ||
| Err(TrySendError::Closed(_)) => { | ||
| // Channel closed - critical failure or shutdown | ||
| error!("QBFT message channel closed - initiating shutdown"); | ||
| // todo!() need some sort of shutdown |
There was a problem hiding this comment.
Sorry, one more nitpick:
warn!(?msg, "Dropping QBFT message due to full queue");And for the Err case: a return after the log should suffice here.
There was a problem hiding this comment.
A bit confused on the Err case. Dont we need some explicit shutdown? Using return here is the same as finishing normally with a implicit (). The instance will still be able to send messages and they will all hit the Closed case. The linter also says the return is unneeded.
There was a problem hiding this comment.
aaah, yeah of course, missed that we are inside a closure.
hm. let's leave it as is for now, the instance also quits if the input queue is closed, so it should be fine
Issue Addressed
#98
Proposed Changes
This PR introduces a framework to provide end to end QBFT Tests. It allows the ability to run multiple instances at once and control the operational status of the nodes along with any byzantine behavior.
There are also a few bugfixes from issues that were discovered during testing.
Notes for clarity
cluster_memberswas changed to aVecas we need deterministic leader selection for testing and random ordering with theHashSetdid not allow for this.