Until I understand what is so horribly wrong, here are my pinned posts: https://pleroma.ledoian.cz/notice/B2fy56ChT14sLKSWkS https://pleroma.ledoian.cz/notice/AkPQ89qoxr8GeUeS3s https://pleroma.ledoian.cz/notice/AjiTdJYO25GMhbIRNY https://pleroma.ledoian.cz/notice/AjRC6uclYTrWz5zV9E
I have no idea yet why the endpoint for pinned posts times out, but it does 🤷
My instance is cursed and wall time should not exist.
I might have spent awful lot of time looking for unanalyzed #postgresql table. The solution was literally two words, I just didn't know :-)
That's what #selfhosting gives you :-D
Wow, I have not seen my instance be this happy. Given how underpowered its hardware has been since the beginning, not seeing database timeouts in the log and having requests be served in few tens of ms even when I post is actually new to me.
Don't worry, the hardware itself is still unstable (bad hardware fix of the laptop the instance is running on), so intermittent issues are still expected :-)
Well, search has given up on me (so not all is too well), so when restoring my pinned posts I had to scroll down my post history.
Two takeaways: (1) I had a good taste in humour even two years ago, and (2) I used to (shit)post a lot here. (Let's see if I can do that again :-P)
I have written a long-ish blogpost about the stuff I needed to do and why in a hopefully more coherent manner: http://blog.ledoian.cz/ipv6-pleroma.html.
If I get to resolve the rest of the issues, I might (or might not) update the blogpost. But the rest is mostly minor issues (unless something big shows up).
Heh, apparently instances of the fourth kind do exist: the ones that have IPv6 outbound connectivity, but no AAAA record so no inbound traffic.
And I see them on my timeline, just can't see media and profile picture.
Now I wonder how to contact the admin in question, given that I am too lazy to create a IPv4-compatible alt account…
I inspected the oban_jobs table out of curiosity, and my planned jobs should mostly run by 2026-02-16. I guess that means that after that, problems on my instance might just become permanent (and thus worthy of being fixed manually instead of waiting and hoping)…
The end to end principle is a lie. It suggests that instance A either can or cannot reach instance B (unidirectionally).
This does not follow my observations – there is at least one instance that I appear to send my posts to¹ (push from me), it managed to get my updated bio (pull from me), meaning that bidirectional communication should likely be possible. But I don't get its posts (push from it)…
It could also be exponential backoff again, maybe. While the other instance is Mastodon, I already know that my Pleroma creates federation jobs per post, not per remote instance, so those timers are independent here. And I have no idea what Mastodon does w.r.t. broken-and-restored federation…
¹The post has just been sent to the public and to my knowledge, hasn't been interacted with. However, Fedi being Fedi, it technically might be also caused by my server not knowing…
Heh, I should have realised earlier: I cannot signed fetch from an instance that cannot¹ connect to my instance, as it does not have a way to get my public key.
Corollary: Even if I follow you, if your server cannot start IPv6 connections, I won't be able to read your follower only posts, even if I learn about them. (Cannot fetch them and you won't push them to my inbox.)
¹More precisely: hasn't been able to connect lately, because of caching