So I finally got tired of the hit and miss of using azorius on the same subdomain as my old lemmy install. I assumed it was a key issue so I killed the DB and re initialized the build, and am now hosting it under the new subdomain deddit.
I subscribed to all the same communities but this time I noticed when loading a new community no posts were shown, though the logs clearly show json dumps with posts from those groups. I didn’t think anything of it and kept subscribing to my groups. Sometimes a couple of posts would show but usually zero posts.
Today, I see a few commeints but no new posts. Even though the groups are active with new posts, etc.
I see a lot of these entries in the logs:
2023/08/24 16:02:45 no post for comment: https://programming.dev/comment/2291160
Anything I can do to troubleshoot this? Are there any key files stored I should have wiped out or something? I did delete the entire -datadir and recreated it.
So is there some like polling that happens after X time? Because I notice I do get new stuff but it’s bursts by groups. So like, the first 10 most recent are from a single group, the next 10 are from another group, etc. It seems like things aren’t being pushed to me but pulled from Azorius.
Yes. Every day or so, we poll group outboxes looking for posts that we missed somehow. This also helps fill in groups that don’t have any subscribers, and would appear dead otherwise.
The code is in refresh.go.
Thanks! I’ll keep an eye on things but it seems I have a delay of up to 24 hours for any new posts in groups I’m subscribed to. I keep referring to world@lemmy.world because it’s a pretty active group but it seems anything pushed to me isn’t sticking. I’ll dig around the logs a bit too and update if I find anything odd.
After a big of digging through logs it seems it’s lemmy.world that is not posting updates to me. Trying to unsubscribe produces this error:
2023/08/29 19:16:01 failed to post json to https://lemmy.world/c/selfhosted/inbox: http post status: 403Are the keys stored in the database (azorius.db) or on disk? I did destroy the db and re-create it when moving to a new sub domain but maybe the keys survived if they’re stored on disk somewhere other than the db? Could that be the issue (though I guess it doesn’t make sense it works with other lemmy instances)
That’s a little weird, but lemmy.world is always having problems. :(
In my testing, a bad key would result in a 400 status.
The keys are stored in the database. Every instance will have new ones.
You may be stuck for a bit in the delivery retry queue. After a failed response like 403, we’ll save that and try it again for a while. But new messages will line up behind the failed message. So effectively, you can’t talk to lemmy.world until the 403 message times out. This is by design, but if there’s a federation problem, it’s not great.
You can peek in the database at the deliveries table, and delete anything with a rcpt of lemmy.world. Then try subscribing again.
Well, I just went and unsubscribed and resubscribed and finally I got
200 OKresponses and Accept posts back. So maybe things are all worked out now. Wish I could say what the issue or resolution was but as the saying goes, time heals all wounds.