🇳🇱
@realcaseyrollins@mstdn.social
I've never tested it, but I would expect thast to be the case. As far as I know you cant effectively load balance a cluster of pleorma instances which itself would be a natural upper limit. There will always come some point where a a single server isnt enough.
@freemo @realcaseyrollins @design_RG @moonman I don't see why it would be a problem to load balance Pleroma. You just add more Pleroma processes and load balance them. One Pleroma service is more efficient than 2 sidekiq processes alone. It's running on a compiled language that supports concurrency. The largest server is My Free Cams. I have no doubt Spinster will have huge cost savings after the move.
Depends largely on how the instances would share session information.
If you are load balancing pleroma instances the implication is the instances are geographically separated running different databases in a cluster formation.
So they would need a way to keep session information synced in real time. Obviously relying n database replication for that would defeat the purpose
Thus why you have things like redis, to share the session information separate from the database itself.
I'm no expert on pleroma but as far as I know there is no mechanism for that. Unless pleroma has implemented it recently or something and either added redis or at the very least stateless tokens to do something similar.
@realcaseyrollins@mstdn.social @design_RG @moonman
Its certainly solvable, and absolutely would only effect the most massive of instances. Outside of M.S I doubt it is even relevant for most servers right now.
This is all a hypothetical in the case of a mass exodus of social media to the fediverse really.
@realcaseyrollins@mstdn.social @design_RG @moonman
As long as there is active development I'm sure youll get there when the time comes.
But my only point was, at the moment, it would appear with a precursory view mastodon is better prepared for massive user count. M.S would be an example of that.
Though I admit it is also mostly supposition, I've never tested where the failure point for a pleorma instance would be or how it might compare at M.S sizes in terms of performance and hardware.
@alex @realcaseyrollins@mstdn.social @design_RG @moonman
@freemo @realcaseyrollins @design_RG @moonman I've never had to scale that far. The larger Mastodon servers I run still have a single database but the web servers are load balanced and on the same network. Either way it sounds like a problem that can be solved in Pleroma with better long-term benefits.