Hi all, I tried hosting my own lemmy instance to take some of the load off lemmy.world, but a lot of the posts do not get synchronised to my instance. When they do, they never got more than a couple upvotes.
I get a lot of warnings in the logs that I’ve got an incoming connection not on my allowed hosts, of which I’ve put “lemmy.world” and “lemmy.ml”
Does self hosting like this actually reduce the load on lemmy.world and the fediverse? And how can I fix the issue I’m having?
You must log in or register to comment.
Leave allowed hosts blank! Otherwise you’ll only get content from those instances.
Also you will only get posts from communities with at least one subscriber on your instance, and only new posts and comments from the moment you subscribe, or if someone interacts with an older post/comment
upvotes are all over the place, posts being out of sync is normal too right now. growing pains.
“Allowed Hosts” is white listing btw. You should leave it blank if you want more than just lemmy.world and lemmy.ml
And I wouldn’t say it reduces the load on those sites. It would kind of add to them in a way if you post from your instance or sub from your instance solo. If it’s just you on your instance, it’s barely a blip. May save them some photo hosting if you post.
SOME of the out of sync upvotes/comments are fixed in 0.18.1
It would kind of add to them in a way if you post from your instance or sub from your instance solo
Is this actually true? As far as I understand it, by far the biggest overhead is users browsing. The fewer users you have actually hitting the frontend, the better.
The devs were saying that browse load was the major issue a few weeks ago, but more instances have been stood up since then and it does seem like replication is getting a bit creaky. I don’t think replication load is flatlining the CPUs of big instances, but it does seem to be reaching some kind of scaling limit that is non-trivial to tune around.
The optimal is definitely the goldilocks zone of ~100 to ~1000 active users. Whether single-user instances are a net win or a net loss is still pretty unclear to me, and if they become a net-loss it will manifest as federation issues of the sort described here, but it’s pretty hard to tell whether those issues are misconfigs on the tiny instances, misconfigs on certain big instances, temporary scaling issues that will be quickly fixed, or hard problems that will limit the size of the network for a while.
