It turns out that running an instance is *not* super easy
@qonnyr backups? high load adjustment? instance reaping? monitoring? paging?
(I worked in a devops/sre team: running a HA service with dynamic load isn't trivial).
@pnathan Not even close to that part yet, heh. Got sucked into a rabbit hole trying to get a Heroku deploy working, because I didn't want to spin up a Digital Ocean droplet. Probably should have. :P
@qonnyr my suggestion is actually more exotic: run an ASG over on Amazon with mastodon fronted by an ELB. when your node dies, the ASG will respawn it. the database can live over on RDS, provided it's not demanding unsupported pg extensions. Redis can be farmed out to AWS as well iirc.
@qonnyr then you can wind up a variety of stats: with sufficient dinking, you can do pretty much all you need.
I'd guess final cost to be ~~~100/mo? 200/mo? Hard to say, I don't have a mastodon instance stood up to estimate load / user + baseline load.
@qonnyr Wow that was *amazingly* incoherent of me, and I've only had one beer.
Rephrased: You can use AWS to drive an *almost* fully HA system with backups, etc, tolerate node death, and include sufficient monitoring and stats that you can drive things like texts when the server becomes unavailable.
@pnathan I really need to migrate my high load projects to AWS at some point. That sounds like a perfect strategy for my CORS proxy project
@qonnyr naw AWS would be way too expensive for crossorigin :P
@tjvr But it might not be down for twelve hours a day then? :P
@qonnyr hit me up if you want to talk about how to do aws right. that knowledge is locked up in horrible consulting blogs or obscurantist docs