Delay considering a channel closed when seeing an on-chain spend #2936
Conversation
|
@thomash-acinq how do you think a splice should change the low/high bounds estimate for a graph edge? This PR currenty increases the 'high' for the edge by the amount a spliced channel's capacity increases and decrease the 'low' for the edge by the amount it's capacity decreases. Is that correct? I tried to copy the logic used currently when an channel is added or removed, but only for the change in capacity instead of the entire channel's capacity. |
That looks good. |
Issue ACINQ#2437 When external channel is spent, add it to the spentChannels list instead of immediately removing it from the graph. Remove spent channels after 12 blocks. When a newly added channel is validated, if it spends the shared output of a recently spent channel then it is a splice. A splice updates the graph edges to preserve balance estimate information in the graph. Channels are immediately removed from the db so they will not added to the graph if a restart occurs before 12-blocks elapse.
remyers
force-pushed
the
2410-delay-close
branch
from
5fbb4a9
to
d1109fa
Compare
See issue #2437
When an external channel is spent, the router will add it to a new
spentChannelsmap instead of immediately removing it from the graph.If after 12 blocks an entry in the
spentChannelsmap is not part of a splice, it will be removed as before.When a newly added channel is validated, if it spends the shared output of a recently spent channel then it is a splice.
A splice will update the
shortChannelIdandcapacityof graph edges of the parent channel instead of removing the parent's edges and adding edges for a new channel. A splice will also reuse and adjust the parent edge's low/high balance bounds information based on the splice.Because a spent channel could be re-added to the graph if our node is restarted before the 12-block delay, spent channels will be removed from the network db at the time they are spent. Nodes that are isolated when their last channel is spent will not be removed from the db until after the 12-block delay. Isolated nodes will also be removed from the network database when our node is restarted.