@hj@lain A quick update, after adjusting postgres config as recommended here the CPU occasionally drops bellow 100% and I even managed to load known network tl without logging in (didn't want to log in to avoid hammering the whole thing by requesting notifications)
Btw speaking about running Pleroma, any ideas what else can be done to drop the CPU usage below 100%? A already deleted deliveries and old notifications, I remember a while ago someone was mentioning clearing oban queue or something like that, how would you do that?
10 is the number of fingers on human hands, not the end-all for database scaling. Having more connections absolutely does help when you have 20K+ users.