Re:Post mortem on mastodon outage with 30k users (hachyderm)
by throwawaymaths on Sunday December 04, 2022 @12:44PM
(#33855818)
Pleroma also isn't RoR and is really designed to scale out better
Re:Post mortem on mastodon outage with 30k users (hachyderm)
by Throwawayaerlei on Sunday December 04, 2022 @01:33PM
(#33856270)
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.
Re:Post mortem on mastodon outage with 30k users (hachyderm)
by zimpenfish on Sunday December 04, 2022 @01:52PM
(#33856445)
[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
Re:Post mortem on mastodon outage with 30k users (hachyderm)
by zimpenfish on Sunday December 04, 2022 @01:49PM
(#33856410)
> 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.
Re:Post mortem on mastodon outage with 30k users (hachyderm)
by throwawaymaths on Sunday December 04, 2022 @06:36PM
(#33859287)
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
I dread to think how much a busy 30k user instance does.