I managed to get it to run. I checked in postgresql logs and it seems it was caused by trying to add into site. Apparently the site name cannot be more than 20 chars but this is mentioned nowhere in the documentation.
And when this happens, it causes this cascade of somehow trying to recreate the admin again and again, which is a red herring after all.
wow very tricky, thanks for reporting back! Can I ask if you’re seeing better performance running postgresql separately?
I don't have something to compare it to. As it's still empty, performance is still good.
Yeah, I encountered this too, but only when deciding to rename the site. When I renamed it through the lemmy UI I was able to find the character limitation and work around it. I bet it has something to do with the federation protocol, and how much memory is allocated to a site's name.
I dropped the contraints manually to test, and it indeed seems to create the admin account 3 times for some reason!
Worse. Restarting the container, seems to keep recreating that user 3 times again. Then always fails with
lemmy_1 | Error: LemmyError { message: None, inner: value too long for type character varying(20), context: "SpanTrace" }
This happened with me because I was adding an admin account explicitly to the hjson file. Once I removed that everything worked.
I was doing that as well, but it wasn't the problem. The problem was apparently caused by a too-long site name
Support / questions about Lemmy.