Lemmy.world is very popular, and one of the largest instances. They do a great job with moderation. There’s a lot of positives with lemmy.world

Recently, over the last month, federation issues have become more and more drastic. Some comments from lemmy.world take days, or never, synchronize with other instances.

The current incarnation of activity pub as implemented in Lemmy has rate issues with a very popular instance. So now lemmy.world is becoming a island. This is bad because it fractures the discussion, and encourages more centralization on Lemmy.world which actually weakens the ability of the federated universe to survive a single instance failing or just turning off.

For the time being, I encourage everyone to post to communities hosted on other instances so that the conversation can be consistently access by people across the entire Fediverse. I don’t think it’s necessary to move your user account, because your client will post to the host instance of a community when you make a comment in that community I believe.

Update: other threads about the delays Great writeup https://lemmy.world/post/13967373

Other people having the same issue: https://lemmy.world/post/15668306 https://aussie.zone/comment/9155614 https://lemmy.world/post/15654553 https://lemmy.world/post/15634599 https://aussie.zone/comment/9103641

  • Admiral Patrick@dubvee.orgM
    8·
    6 months ago

    For that example, there may be something else going on. The time behind graph for Lemmy World- > Hacker Talks seems to only average about 90 seconds. Some of the comments that show up on LW were made hours ago. If it were just federation delay, they should have arrived already.

    Assuming those are going straight to your Lemmy server and not to a proxy that buffers them? If you’ve got a federation proxy buffering those, then the graph would be off since it wouldn’t show what’s sitting queued in your buffer.

    You still definitely have a point. Every couple of weeks, my instance seems to start lagging behind by several hours and then goes back to normal after a few days. Still not sure why unless there’s just a huge uptick in activities being sent out from LW. I made a post about it yesterday, but it doesn’t dive too deep into the technical details; basically just an info post to let people on my instance know I’m aware and trying to do what I can about it.

    • jet@hackertalks.comOP
      3·
      6 months ago

      Good point. And thank you for the graph. It could be that the activity pub protocol is best effort, and if lemmyworld is overloaded, or just dropping packets, whatever activity was supposed to be sent at that moment gets lost forever. That would explain me divergence, and lack of eventual consistency

      But I only have data for missing posts from lemmy world, and only some communities.

      I don’t know the actual cause of the issue, but I think it is definitely related to the size of the instance. So from an empirical perspective the advice is still promote communities on non-mega instances