Everyone has an opinion, and at the end of the day, whatever works best for you is what you should stick with.

I like Traefik because you can mount /var/run/docker.sock:/var/run/docker.sock:ro to Traefik, then it can read labels from containers, and automatically wire up new instances based on labels on them. I’m sure there’s equivalent in other reverse proxy solutions, but as I said, it works for me and I like it.

I give that container my Cloudflare origin certificate, everything gets encrypted in transit to Cloudflare, and then Cloudflare handles all the SSL management for me, as well as provide extra layer of DDOS protection.

Yeah, I’m getting mixed results as well. Federation seems to be super finicky right now. A lot of finger pointing going on and some posts I’ve seen suggests it is Cloudflare being the culprit. As much as I’d like to shed Cloudflare to get federation working, I just don’t see that being something that’s viable long term. It is very easy to DDOS someone, and I do not want to expose my instance IP publicly.

Looking at the commit logs, the difference between 0.17.3 and 0.17.4 seems to be just some database optimizations, so I think the problem we’re seeing is still something else.

Also, the lemmy.ml instance is acting up across the board, even from the lemmy.world instance, or other major instances, the subscribe doesn’t seem to return properly… so I wouldn’t necessarily use them as the benchmark.

Yeah, the entire setup is quite finicky still. Part of me thinks Fediverse is forced into the spotlight by Twitter (Mastadon) and Reddit (Lemmy), and the whole thing is not quite baked yet. Don’t get me wrong, having a more open space is great, but there are so many things that’s not quite ready for prime time. I hope the dev team behind the platform (not the self hosted instance admins) will be more open to ideas and rapidly improve the platform.

I am new to the fediverse, and I don’t use Friendica, so I could be entirely wrong about this. However, from what is described, perhaps Friendica has some sort of feature in which would trigger your instance to go out to fetch some data from another instance. Someone exploited this feature, spammed your instance with content from assortment of subdomains on the *.activitypub-troll.cf domain, and most if not all of them are probably non-existent. As result of that, your server is re-checking every 10 minutes to see if they’ve came back online. This would also explain why shutting down the Friendica service resolved the problem for you.

Are there ways to manage lists of such? For example, on the former platform that doesn’t deserve a call out, you can do “me_irl+meirl” and aggregate both into one feed. This makes reading the (albeit potentially cross posted) content in a unified feed much easier.

Another similar point I’m having a hard time getting over is that with a centralized platform, it is easy to go to “Subject A”, and see everything on that subject. However, now I need to see “Subject A@lemmy.world”, “Subject A@lemmy.ml”, “Subject A@someother.instance”… Yes, I could subscribe to them all, but this ultimately end up creating a noisy home feed with also “Subject B@lemmy.world”, “Subject B@lemmy.ml”, “Subject C@lemmy.world”, “Subject D@lemmy.ca”, … etc. all baked into one feed, as opposed to just something focused on “Subject A”.

Lastly, discoverability leaves a lot of room for desire. Today, I’m fairly new to Lemmy, I am actively seeking out communities that I might be interested in, across multiple popular instances, and hoping that federation is enabled between the two instances. Tomorrow, I’d find that I’m subscribed to too many (see the noisy main feed issue above), and I’d remove a bunch. Next week, am I likely to go to the Join Lemmy directory to find new instances, and add “duplicate” communities from newly popular instances? I think not.

I think the long term survival of the platform (to expand beyond just us tech nerds that hate the former platform) will depend a lot on streamlining this workflow to make content discovery much more consistent. Even a simple option where a pseudo “!Community@” (with no instance) feed that aggregates all the “!Community” regardless of instance that you’ve subscribed to, might go a long way.