Node
Node copied to clipboard
MASQ-Project
Inspect code for handling a dropped TCP connection between Nodes
Relates to graceful shutdown, but this card context is around if an active TCP connection is broken between Nodes.
Need to inspect the code to see if this is handled in any way
If the code exists, its likely not being actively used by masq
I believe there used to be a card considering an implementation of some sort of Ping message to be transmitted every so often trough the maintained TCP connections to all the immediate neighbours to assure the transmitting Node that these supposed neighbours are still up and healthy.
I remember this just vaguely but I think we concluded after a search that we cannot find out about the connections being broken until we try to actively use it for something (classic traffic or Gossip) and we fail. Then, I think, there is some code to purge this Node, so that next time we won't consider it anymore because the Node will be neither in the Neighbourhood database nor the stream pool with kept handles to (rightly functioning only) TCP connections.
But as it goes with me I might be wrong about a few things here, so it is advised not to take this an unquestionable truth, rather an impulse to refresh half-forgotten ideas.
With the GracefulShutdown, one part says we need to convey our leave to our Neighbours upon which they will react by removing us from their database and the stream pool. That will probably eliminate the most frequent situations of confusion about Nodes that have been intentionally ordered to shut down but without awareness of their immediate Neighbours discovering it some time later indirectly when failing to communicate with the neighbour being gone already.
Other situations might be, though, that the other Nodes panicked and there were might be asking if well it would be worth it having this periodically outgoing Ping message to know about it.
If it comes only to panics, that we obviously do not expect to unwind often, it might not be so important anymore, because we will always eventually realise the other Node's absence when we first try using it again for communication. Is that dirty? I don't know, it seems legit to me.
I found the card mentioned in my first comment. You can reach it under the name '"Ping" operation for network links', or GH-705 in the old board ordering.