Skip to content

Comment on Post mortem on Mastodon outage with 30k usersparent

Comments

Looking at the amount of requests my 5 user (following ~60 people in total), 90% idle Pleroma instance handles is a bit mad - 60-90 requests a minute.

I dread to think how much a busy 30k user instance does.

Pleroma also isn't RoR and is really designed to scale out better

Pleroma also isn't RoR and is really designed to scale out better

Sure ... but that's not what I was talking about. It's still talking ActivityPub and that seems, to me, from my low use, few user instance, to be an extremely chatty protocol. I don't know if the ActivityPub traffic scales linearly with users but it would be a not inconsiderable number for 30k users.

Exactly. We're agreeing! That's why I'm not surprised that the mastodon server didn't do so well and that you might get more mileage out of your compute

My understanding is that Pleroma was first and foremost designed to have a small memory footprint for small instances. As an Elixir program running on the BEAM (Erlang) runtime it ought to scale a lot better, but serious work had to be put into that. Not sure it's running on any sites as big as the big Mastodon ones.

[1] has the biggest Pleroma at about 27k users - about 45th biggest according to the list on [2]

[1] https://the-federation.info/pleroma [2] https://instances.social

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.