[bug]: Waiting Close Channel - ChansStatusDataLoss

[sperolightning]

Background

it happened that there was a power failure and subsequently it appeared that there was a "waiting close channel". The node is synchronized and I tried to perform the forced closure command as read in other posts:
lncli closechannel – force “channel_points” 1
but the following message appears
rpc error appears: code = Unknown desc = cannot close channel with state: ChanStatusLocalDataLoss
Other user resolved by contacting the other node forcing closure. In this case the node is Kraken, and I imagine it is not easy to contact and solve the issue. I wonder if there is a software solution to solve the problem?
Thanks for your advices

Your environment

Umbrel Version 0.5.3 on a Raspberry 4b 8GB HDD 2TB.

Steps to reproduce

lncli closechannel – force “channel_points” 1

Expected behaviour

closure of channel

Actual behaviour

rpc error appears: code = Unknown desc = cannot close channel with state: ChanStatusLocalDataLoss

[guggero]

It sounds to me like the power outage caused data corruption in the channel DB and lnd noticed on startup that its commitment height didn't match the remote node's height and went into safe mode for that channel (ChanStatusLocalDataLoss).
The following should work:

1. Make sure you have a backup of your channel.backup file on another machine than just your Umbrel. Creating a backup of the channel.db file might also be a good idea here in case things go sideways (just never restore that file to its original location, only use it with chantools when necessary).
2. Export the channel backup for the channel in question: lncli exportchanbackup --chanpoint <xxxx:y> --output_file /tmp/chan1.backup
3. Try restoring that single channel from backup: lncli restorechanbackup --single_backup /tmp/chan1.backup.
4. Check the log for the channel in question. If it says something like "already exists in channel DB" then you need to remove it from the DB first:
   1. Shut down lnd and make sure it stays shut down and isn't just restarted by Umbrel (might need to consult their docs to figure out how to do that).
   2. Run chantools removechannel --channel <xxxx:y> --channeldb <your_channel_db>
   3. Run lncli exportchanbackup --chanpoint <xxxx:y> --output_file /tmp/chan1.backup and check the logs.

The above should ask the remote node to force close the channel from their end.

But in cases with data corruption to the channel.DB I recommend to start with a fresh DB just to make sure there aren't other areas affected. If you decide to go with that option, you can simplify the steps somehow. Just close all normally operating channels cooperatively (that should save on fees compared to using the backup), then stop lnd and remove the channel.db, restart and use the full channel.backup file to get the remaining, borked channel back.

[guggero]

I'm also going to convert this to a discussion as this is the current expected behavior of lnd if it detects data corruption. We are actively working on migrating to another DB format that is less prone to corrupting on sudden power loss.