Mastohost @mastohost is taking on new accounts again. You can get your own Mastodon server for $6 a month! The great thing about it is they take care of the software, you do not need to be technical to run a server.

Mastohost is where sauropods.win is hosted, and I can recommend them. Support is top-notch, and it's all been smooth sailing.

https://masto.host/

#hosting #mastodon #fediverse

Masto.host - Fully Managed Mastodon Hosting

Masto.host was built from the ground up to make running a Mastodon instance easy.

Masto.host
@john @mastohost realistic question is how many people can you even host on $6/mo, storage, etc?

@husbandpanda It depends on multiple factors. I have people on the $6/month plan with having several dozens of active users. I have people on the Galaxy plan with a single user.

If you follow or are followed by thousands and thousands of remote user then the $6/month plan will not work. If you have a group of friends that almost don't follow or are followed by anyone remote, you can probably have a couple dozen without a problem. Again, it depends on the users you are hosting.

@john

@mastohost I can understand that in general, but to avoid oversimplifying and delve a little - the server capacity limitation sits within what? Storage? I/O? CPU? Memory? What gets taxed with (dozens of active users) vs (tons of cross server follows)?
@husbandpanda It's hard to say. Per hour, how many posts will the active users make? How many of those posts will include media attachments? How many of those attachments are videos? How many interactions (boosts, favs, replies) will the users make? The answer to each of those questions will have a different impact in terms of CPU/IO/RAM etc.
As usually one cannot predict that, it's hard to determine and I don't even track/limit things like CPU/IO/RAM per instance.

@mastohost so basically $6 may or may not fit whatever and nobody's #monitoring and hopefully things are fine? lol. Sounds like a long term risk.

As a monitoring engineer by profession I would hope people are monitoring things on some level if it's their instances. Not to criticize but just in general for your own sanity.

I get that social media is what it is and people do whatever but it shouldn't be impossible to quantify tbh. Maybe I need to make this a Rust project I do.

@husbandpanda @mastohost *weeps in observability at the idea that you just need to ship metrics*
@markdennehy @mastohost right now it sounds like there is literally 0 though, so we can't even figure out what might need to be observed. Obviously qualitative analysis of what metrics can be found is kinda key but....
@husbandpanda @mastohost I mean, that approach is valid for the scientific study of previously unknown things, or sherlock holmes investigating something, but it's not what you should do in this context. Someone wrote the parts involved, someone knows how they string together, someone knows where the bottlenecks are (the .ie instance got tuned when Irish twitter first decamped so the knowledge exists). That means we have a rough system model and can identify the important things to monitor.
@husbandpanda @mastohost Find the admins who've had to do performance tuning, ask them what they looked at, what they thought the important parameters were, instrument those by default. Since there are lots of similar mastodon instances, you'd be doing a favour to a lot of new admins because it'd be widely applicable.

@husbandpanda As far as I know, nobody has been able to determine CPU/RAM/IO usage needs based on the number of active users.

It's pretty much the same as you being able to determine the CPU/RAM/IO of a WordPress blog, based on the number of authors. I don't think one can determine that.

@markdennehy

@mastohost @husbandpanda I think that sounds more like provisioning and yeah, that's way harder because you'd have to both know how many users and what they're going to be doing. The stuff the .ie instance went through seemed more like people tuning stuff not based on users but on queue lengths and other internals, after the users had all landed and the strain had started breaking stuff.
That's more like shoring up a bridge by what bits break instead of counting cars.
@mastohost @husbandpanda Still, it'd give you a list of what bits broke, and what bits you shored up. That's a pretty good model.