Channel balance_msat becomes negative

[cfromknecht]

In lightningnetwork/lnd#1499, @mojoaxel provided lnd and c-lightning logs related to a channel closure. I haven't looked into why it was force closed, but i did notice in those logs that the channel's balance_mast becomes negative. Here's the relevant comment: lightningnetwork/lnd#1499 (comment).

I'm not sure if there is a deeper issue or what it might be, but an extra assert(balance_msat >= 0); inserted here would probably help track it down. Extensive c-lightning logs were also provided in the issue, hopefully they're helpful to you!

[rustyrussell]

Thanks for spotting!

It confused me to, but it's actually just a debug msg error. We're (correctly) rejecting the local attempts to add an HTLC precisely because it would drive us under the reserve value.

Still owe you a beer for reporting the error :)

[rustyrussell]

BTW, we're getting upset with (lnd-looking) peers for this:

2018-07-05T05:11:25.881Z lightning_gossipd(1690): peer 02a54deb8d0f11d47c6f55cec5e673063c9fad2619559e8d87ae3eb4c381668449: Bad concurrent query_short_channel_ids

You're supposed to wait for a reply before asking again: known issue?

[cfromknecht]

@rustyrussell glad to hear! no bugs is always good :)

re: gossip queries, i believe that issue is fixed in master now. We would try to query for premature anns, which I'm guessing would overlap with the gossiperr syncer state machine that's maintained separately
https://github.com/lightningnetwork/lnd/pull/1798/files

[cdecker]

@rustyrussell that commit isn't in a PR yet, is it? I haven't seen it so far and it seems like a simple fix that we could merge immediately.