Open geigerzaehler opened 3 years ago
This isn't necessarily an incorrect termination as I understand things. The local peer is trying to send a Disconnect
to the remote peer but the remote peer has already gone away. There are a bunch of ways that could happen I think. For example, the remote peer could be spamming unsolicited messages, then being rate limited and therefore dropping the connection on it's end before we're able to send the disconnect.
Yes the peer is obviously already disconnected (NotConnected
). It's a bit funky that there seems to be a backlog of broadcast messages still being applied. Not sure there's much we can do about this, but perhaps rate-limiting the egress disconnects would be a good idea, so we just drop the message on the floor.
I'm not sure which codebase this is in these days, but I get the impression something is pretty wrong with those wants (it's always the same object they're requesting).
When a peer sends unsolicited messages the connection to the peer is not properly terminated under some circumstances.
This was witnessed in https://github.com/radicle-dev/radicle-upstream/issues/2010. There, you can also find the complete logs.