[bug]: 0.19.1 filling logs with: [WRN] GRDB: Channel=969519666583044097 not found in graph cache

[Dominion5254]

Background

After updating from 0.19.0 -> 0.19.1 the logs seem to be spamming the below warnings despite lncli getinfo showing "synced_to_graph": true,

025-06-12T10:44:12-06:00  2025-06-12 16:44:12.543 [WRN] GRDB: Channel=969519666583044097 not found in graph cache
2025-06-12T10:44:12-06:00  2025-06-12 16:44:12.543 [WRN] GRDB: Channel=969519666583044097 not found in graph cache
2025-06-12T10:44:32-06:00  2025-06-12 16:44:32.515 [WRN] GRDB: Channel=933411704926044160 not found in graph cache
2025-06-12T10:44:32-06:00  2025-06-12 16:44:32.516 [WRN] GRDB: Channel=933411704926044160 not found in graph cache
2025-06-12T10:44:43-06:00  2025-06-12 16:44:43.369 [WRN] GRDB: Channel=989992573135683585 not found in graph cache
2025-06-12T10:44:43-06:00  2025-06-12 16:44:43.369 [WRN] GRDB: Channel=989992573135683585 not found in graph cache

Is this expected to resolve with more time?

LND seems to otherwise be functioning normally.

Your environment

• version of lnd: 0.19.1
• which operating system (uname -a on *Nix): StartOS 0.3.5.1
• version of btcd, bitcoind, or other backend: bitcoind 28.1

Steps to reproduce

Update LND from 0.19.0 -> 0.19.1

Expected behaviour

Clean logs without these warnings when the graph should be fully synced

Actual behaviour

LND logs spamming warnings about channels missing from the graph cache

[wereHamster]

I don't think this is a new issue, I've seen this warning even before the upgrade from 0.19.0 to 0.19.1.

#9930.

[Dominion5254]

It is definitely possible this issue exists on 0.19.0 (which I admittedly hadn't used for very long before updating to 0.19.1); But I can't said I have seen this warning on LND <= 0.18.5.

[wereHamster]

The log was added in 0.19.0 (001e559). Though the problem likely existed before (just silent).

Is it possible that you were running with Neutrino backend initially and then switched to btcd / bitcoind backend?

[Dominion5254]

Nice find. I guess the question is whether LL intended for this additional logging.

I am not nor have ever run neutrino on this node.

[saubyk]

Converting this into a discussion thread instead

[wereHamster]

Channels are not added to the graph cache when both policies have it disabled, which is a realistic scenario. Then this message is generated later on when somebody tries to lookup this channel.

https://github.com/lightningnetwork/lnd/blob/fe405426cac5096773cc039dff1cb20367aeb54d/graph/db/graph_cache.go#L124C2-L128C3

[whizz]

I am seeing quite a lot of these messages too. When I query lnd by using lncli getchaninfo, I can see the channel is known and indeed disabled on both ends. I'd say this message is not that usable. And it is always included twice (because the cache lookup is done twice for node1 and node2.

[remcoros]

Any update? My logs are full of these messages and it's very spammy. It's not really a helpful message either. Is this really a warning? Should it not be a debug log?

[lnliz]

same problem, lot of log spam fill up logs
(not running on neutrino backed)

[ViktorT-11]

@remcoros & @lnliz, are either of you running with an SQL db.backend, or are you running with a bbolt database backend?

[remcoros]

bbolt backend.

the following PR added the warning log (it didn't log anything before): https://github.com/lightningnetwork/lnd/pull/9424/changes#diff-d399a1945b78e9705eb9e307dddadfc854146e3a75c8fd5e935aedba31437f2cR249

It looks like, if this is a common case and if it should be logged at all, it should not be a warning, but a debug/trace log.

[lnliz]

bbolt

[ViktorT-11]

Thanks @remcoros & @lnliz. We've now merged a fix for this: #10514
We'll aim to get that into lnd v0.20.1-beta.