this post was submitted on 22 Aug 2023
3 points (100.0% liked)
Lemmy Support
4656 readers
18 users here now
Support / questions about Lemmy.
founded 5 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
GitHub issue about this topic: https://github.com/LemmyNet/lemmy/issues/3782
Some of it might be avoided by tweaking the PostgreSQL database with higher key values after the new install, but the whole situation isn't really planned for or recognized in the code
Might it be a good idea to to just copy the post so it gets a higher ID, which wasn't used before? Because the productive system is already running and has a lot of content and a new reinstall might even block more IDs.
if it were me right now with Lemmy 0.18.4, I'd take the server offline, do a PostgreSQL dump file - keep a copy, then hand-edit the sequence numbers in the dump file - and do a restore.
you probably only had a few users, so I would set user to 100, person id can be higher because of federation - but jump ahead to 10000 maybe. Post and comment set ahead to 10000 ... and community set ahead to 10000 because that gets federated
the PostgreSQL sequence numbers should only get used on newly created objects here-forward.
That sounds complicated and I am afraid something could break (again). The last days I tried to fix something in the db, which led to other problems.
It is complicated. It's surely a damned-if-do damned-if-don't situation. It doesn't sound like you had all that much in terms of local users, communities, posts, comments - so at least that's in your favor.
I tried to copy some posts, but it looks like it is totally screwed up, now. Even higher ids show wrong content on some Mastodon instances, even if I am sure I only did a few tests with posts.
Only way to solve this (imho) is to reinstall Lemmy BUT use another subdomain.
What do you think? Will this work?
I wold agree that this is worth considering as an approach to not clash identity and get into custom SQL or Rust programming. But there isn't even really a procedure in place to decommission the old lemmy entity... so another damned if you do, damned if you don't in 0.18.4 era.
I'm a little surprised that the federation private key/public key signing doesn't get upset about all new keys appearing on the same domain name. I've tried to get details of exactly how a server joins the Lemmy network and gets discovered over on !lemmyfederation@lemmy.ml but haven't gotten any actually discussion on the details.
I've seen people nuke and start-over their database from empty several times while having problems setting up NGinx and Docker... or whatever part.
I'm glancing at the list of SEQUENCE in Lemmy....