I have posts and comments on Hexbear from 4 hrs ago that don’t show up

I have a comment on Lemmygrad from 4 hrs ago that doesn’t show up

A recent series of posts from ml to other instances have almost no interaction locally. When viewed remotely, they have higher vote counts (no more comments to compare).

I haven’t extensively looked at the new feed to see if posts are outright missing, but I see the same general pattern spot checking posts that do appear: vote totals much lower than elsewhere, missing comments

I don’t see issues reported on https://phiresky.github.io/lemmy-federation-state/site

  • Muehe@lemmy.ml
    link
    fedilink
    arrow-up
    1
    ·
    8 days ago

    Hmm, yeah, the post I linked above is still missing some 50 comments, and most of the post and comment upvotes. Kind of interesting though that it acquired 92/148 comments (at time of writing), before you pushed the fix yesterday it had 2 upvotes and 0 comments. Guess the upvotes are just what has been added since then.

    Would be nice if there was a contingency option for server admins for situations like this, no? Something like “Resynch federation events from <timestamp> to <timestamp>”, maybe limitable to certain instances/communities/posts. Not sure if that’s possible with how the federation works currently, but if it is it would be a good feature I guess.

    • Nutomic@lemmy.mlM
      link
      fedilink
      arrow-up
      2
      ·
      7 days ago

      Normally Lemmy returns an error if activities cant be handled (eg instance is down). Then the sending instance keeps retrying until it is successful. But in this case its a bit different, and the sending looked successful.

      Anyway I made a change to the federation library to add Host header automatically if missing, then this problem wont happen anymore in the future.

      • Muehe@lemmy.ml
        link
        fedilink
        arrow-up
        2
        ·
        6 days ago

        But in this case its a bit different, and the sending looked successful.

        Yeah I got that. But couldn’t you (theoretically) add a “resend activities from <timeframe>” option to the API? So you can ask other servers for things you missed. I mean what happened here is one problem where that may be useful, but another would be something like having to restore from day old backups, being a new instance with no history, a network outage, or missing things for whatever other reason.

        Like I said I don’t know the data structures involved, but in my layperson head it sounds possible, although you may have to throttle it to prevent abuse I guess.