Pulling on a thread context here. We are on rke2 1.33.4 and running the version of kine that is bundled with it. The spammy error message we see in the log is

Oct 28 16:08:49 notice k3s: time="2025-10-28T16:08:49+01:00" level=error msg="TTL delete trigger failed for key=/registry/events/<podname>-59dd494c99-6zccx.18716e967027d1c1: context canceled, requeuing"

While this appears to happen a lot for

/registry/events

in our case. This appears to happen with any data stored in the sqlite database. When we inspect the database with the following command

sudo sqlite3 state.db 'SELECT * FROM kine WHERE name LIKE "/registry/events/<namespace>/<podname>%"'

I expect to see entries for

59dd494c99-6zccx

. However we don't, and it appears that kine is never able to recover, and will continually attempt to delete a key that is not in the database.

the only time the main context is cancelled is when kine as a whole is getting shut down so I’m not sure what would be going on there… but it probably shouldn’t re-enqueue if the error is related to context anyways

@creamy-pencil-82913 thanks for the reply, can you think of any other steps that I could run through if I get a node in this step to help dig deeper?

I think I found the bug… I suspect it’s coming from the temporary kine that is started up initially to extract bootstrap data. The workqueue is not getting shut down properly.

This makes a lot of sense, we were seeing this a lot when rebooting, while moving to newer versions.