it's most likely coming from postgres, db server tends to be the main resource hog.
reinar
joined 1 year ago
Yes, big instances are not ready to handle the traffic and could go down - see lemmy.ml
Are all these thousands of lemmy servers useless?
almost. It's actually worse than that - when you subscribe to a community from your server it will fetch like 20 posts and that's it, you'll get only new stuff after that, so there's no possibility to do a full mirror of selfhosted, for example, if you started your instance today and didn't fetch posts and comments manually.
ActivityPub per se is just a spec on s2s/s2c communication, which is not a great thing since in many cases it assumes single source of truth, which potentially puts huge load on more popular instances.
I think a quick and dirty hack to this could be the following - each linked instance may maintain cache of announces (so there would be benefit of just forwarding original http signed requests w/o being afraid of malicious actor), which your instance could pull, this way you could populate your mirror without overloading the original source.
Distributed activities propagation though... Let's say there are some design steps involved to make this truly distributed, however I feel like it's possible.
https://mullvad.net/en/blog/2023/5/29/removing-the-support-for-forwarded-ports/
They already disabled the possibility to add new port forwards and will remove any existing ones completely on 1st of July.