Processing asynchronously received messages.

[egonspace]

Summary

https://github.com/line/ostracon/blob/dbe53cb7537246681e2d87de51000c02ece2c49a/p2p/peer.go#L408
When the channel of the reactor is full due to the above code, messages are discarded. In the case of a mempool reactor, if a message is discarded, there may be a problem that tx remains in the sender's mempool. When the channel is full, consider waiting rather than discarding the message.

Problem Definition

Proposal

---

For Admin Use

• Not duplicate issue
• Appropriate labels applied
• Appropriate contributors tagged
• Contributor assigned/self-assigned

[tnasu]

When the channel is full, consider waiting rather than discarding the message.

Totally, I agree with you.

I'm not sure, do you mean to change all asynchronous reactor behavior? Remove the default clause?

• mempool/reactor
• evidence/reactor
• blockchain/v[0,1]/reactor
• consensus/reactor
• statesync/reactor
• p2p/pex/pex_reactor

If we allow blocking in receiving messages, there is no problem. But the behavior becomes synchronously reactors. If a node receives too many transactions and blocking the transactions, then the node also cannot work to do consensus/statesync/etc. We should improve behavior to take into consideration synchronously each reactor's behavior. Or do we simply allow becoming synchronously reactors?

[egonspace]

I think it's meaningful because it works synchronously only when the channel is full, and before that, it works async. It would be better to act the same for all reactors.

[tnasu]

OK, let's remove the default clause.