funding: notify on pending channel timeout #7228
Conversation
After a long timeout, the non-initiating node should abandon the pending channel to save resources. The initiating node, however, needs to remember the channel since their funds are at risk.
If we previously notified on pending open, we also need to notify if we close the pending channel. Otherwise, things like channel backup won't be aware of the change until LND is restarted.
When we abandon and close a pending channel, we should also tell the ChainNotifier so it doesn't keep watching for onchain events related to the channel.
|
@saubyk Do we want this for 0.16.0 or 0.16.1? |
Tagged the issue for 0.16.1. Scope for 0.16.0 is now frozen |
|
Some new itests are failing on this PR, and I can't repro locally. Marking this as a draft while I debug. I may push some debug patches or speculative fixes. |
Ensure active nodes are synced to the latest block mined, since the nodes may actually be "fully synced" to an old block when SyncedToChain is true.
| // After more than 2016 blocks without the funding transaction appearing | ||
| // onchain, Bob should forget the pending channel as non-initiator while | ||
| // Alice remembers it as initiator. | ||
| ht.MineBlocksAndAssertNumTxes(2100, 0) |
There was a problem hiding this comment.
hmmm that's a lot of blocks to be mined in a single test...
There was a problem hiding this comment.
We can probably disable the test by default, since we have unit tests for the same thing.
The reason for the itest was to reproduce the log spam from #7208 and verify the fix.
There was a problem hiding this comment.
I think it's nice to have an itest covering it. We could probably make this value smaller for itest so it's easier to test. Or, maybe recreate miners after a certain block height?
Roasbeef
added
funding
For the non-initiator funding flow, the initiator is responsible for broadcasting the funding transaction. If the initiator fails to broadcast, we currently time out the pending channel after 2016 blocks.
However, we aren't notifying anyone else about the timeout, so things like channel backups and the channel arbitrator are continuing to operate as if the channel was still pending.
This PR notifies other components on timeout, allowing them to clean up their state properly.
Tested:
Fixes #7208.