Avoid force-close: ping peer before sending HTLC

[C-Otto]

Background

My node often (21x since 08/Feb/2022) force-closes channels because my peer didn't settle/fail a HTLC in time, where this HTLC is only included in my peer's force close transaction (remote commitment).

I believe this happens if my node sends out a HTLC, but my peer simply doesn't respond. As such I believe it would help (in most, but not all cases) to ping my peer and request a pong response before sending out a signed commitment with the new HTLC (update_add_HTLC and commitment_signed?).

The corresponding message:

[INF] CNCT: ChannelArbitrator(CHANPOINT): immediately failing htlc=HTLCID from remote commitment

Before:

[DBG] HSWC: ChannelLink(CHANPOINT): queueing keystone of ADD open circuit: (Chan ID=..., HTLC ID=...)->(Chan ID=..., HTLC ID=...)
[ERR] HSWC: ChannelLink(CHANPOINT): failing link: unable to complete dance with error: remote unresponsive

Background:
https://bitcoin.stackexchange.com/q/113335/7214

Your environment

• version of lnd: 0.14.3-beta-rc1

• which operating system (uname -a on *Nix): Linux server 5.10.0-10-amd64 #1 SMP Debian 5.10.84-1 (2021-12-08) x86_64 GNU/Linux

• version of btcd, bitcoind, or other backend: bitcoind v23

Steps to reproduce

Send HTLC to peer. Shutdown peer after receiving my signed transaction but before sending back the response (which is the force-close transaction including the HTLC, signed by my peer, that I could use to force-close the channel).

Expected behaviour

I'd expect lnd to not send the HTLC to a peer that is unresponsive.

Actual behaviour

lnd sends out the HTLC and, once it times out, force-closes the channel.

[Roasbeef]

Closing as duplicate of #2992

[Roasbeef]

Also FWIW, the remote peer being unresponsive isn't what caused the channel to force close in this case. This isn't an error we'll force close on.

[C-Otto]

Are you 100% sure about this? I see more than one example where a force-close happened just after this message. The code also looks like this is the case, and I see plenty of good reasons to close a channel so that my peer can't settle the HTLC in the far future.

[C-Otto]

Here's the code that determines if we need to go on chain:
https://github.com/lightningnetwork/lnd/blob/master/contractcourt/channel_arbitrator.go#L1518

Invoked from here:
https://github.com/lightningnetwork/lnd/blob/master/contractcourt/channel_arbitrator.go#L1804

[Roasbeef]

Are you 100% sure about this?

I was referring to the fact that us sending that error won't trigger an auto shutdown. However if for some reason we aren't always cancelling back that HTLC (w/ mailbox auto expiry) then a force close may be resulting. We're tracking that in #6485