Stuck transactions after initial send about lightning

[cdecker]

After discussing this issue on IRC we came to the conclusion that the 0218 node was restarted and it chose a different random port, making re-connecting impossible. Maybe we could write the port back into the config if we have chosen it randomly so that we at least always restart with the same port?

[rustyrussell]

Christian Decker [email protected] writes:

After discussing this issue on IRC we came to the conclusion that the
0218 node was restarted and it chose a different random port, making
re-connecting impossible. Maybe we could write the port back into the
config if we have chosen it randomly so that we at least always
restart with the same port?

Rewriting config files is usually bad. I would prefer that we don't
create the listening at all unless they specify a port in their config?

Cheers,
Rusty.

[cdecker]

I'm fine with that for now, with 0.6 we will also have a default port, so we won't have too many people not accepting incoming connections :-)

[tdaede]

I imagine the P2P discovery will eventually allow rediscovery when an IP or port changes. IP addresses could also change - it would be nice to have an "addnode" RPC call to update the peer address table.

The other problem is that it wasn't obvious what was making it stuck, but now I know to check both getpeers and getnodes. Might be nice for lightning-pay to do the same.

[cdecker]

As soon as we have the new routing announcement protocol from #88 I'll also update existing channels so that reconnecting works.