Figure out why connections are dropping

[germsvel]

When we test our networking layer by spinning up a parity node, either parity drops the connection at some point after the encrypted handshake (sometimes we exchange Hellos and Status) or we throw an error for "Failed to match header ingress mac". Which failure comes up first seems somewhat intermittent.

We are unsure why this is happening. It is possible that it is due to some of the things we have not yet implemented (such as RLPx control flow #408 or handling multi-frame packets #97) or it is possible we just have an error in how we're processing the secrets, since the ingress and egress macs need to be continuously updated for message authentication.

How to reproduce this

The version of parity I am running is 1.9.7-stable

It is best to spin up a parity node without discovery and without other peers. Run,

parity --chain=ropsten --no-discovery --reserved-only --logging=debug

Get the enode url from the parity logs.

In mana, cd into apps/ex_wire. Type,

iex -S mix 

Then from the iex,

ExWire.PeerSupervisor.start_child("enode://url_you_got_from_parity@host:port")

That should start the communication between the mana node and the parity node.

Example

An example of the communication between the two nodes can be seen below:

mana

parity

[InoMurko]

Disk logging which would allow us to review all possible reasons why connections exit are being added in #409 and manipulating dropped connections worked on #688.