On average, it looks to be less than 2gb of ram at the moment. CPU and RAM usage obviously will go up as I have more users, but it’s not bad at all at the moment. I’ve been pleasantly surprised tbh. I am also completely prepared to scale the server up if I get more users on my instance.
Edit: just a follow up, looks like I can scale my instance to a maximum two ways,
“cpu optimized” up to 48 vCPU and 96gb of ram
“Memory optimized” up to 32 vCPU and 256gb of ram
I’m a long way off of the max though now, my server is only 2 vCPU and 4gb memory for now
I’m running a lemmy instance and using about 700mb, up from 500mb before I had any users (though I have maybe a dozen active users lmao)
But I’m not using much CPU at all though. 5% average on a 2core VPS VM. 4 gigs as well. I can scale up a bit and still afford it personally. After that Ill have to ask for donations, and if not enough stop registration.
The scaling is from my cloud provider, hopefully I won’t have to scale up to the max (looks like it’d be like $1300/mo)
700mb of ram isn’t bad at all. Yeah I’m using like 30% cpu on a 2 core right now. Kbin definitely uses more resources than lemmy but I think it has a lot more going on in the tech stack
Yeah if my instance got busy enough that would be something to definitely consider. I’m using DigitalOcean over something like AWS as well. I wouldn’t want to run it out of my home off of my own network though which is why VPS providers are nice (plus I have all my DNS rules set up through it)
I’ve never used a colo before; my rack servers live happily in my basement. If you google say “1U network colo <city name>” you’ll find many providers.
Looking at the recent Docker Compose commits, kbin should scale horizontally until it hits limits of postgres.
Its a really good candidate for kubernetes, if you deploy on AWS/Azure and use AKS/EKS with Azure Database/RDS you will be able to flexibly scale far beyond those limits.
I have been meaning to learn Helm for ages. This seems a good excuse.
On average, it looks to be less than 2gb of ram at the moment. CPU and RAM usage obviously will go up as I have more users, but it’s not bad at all at the moment. I’ve been pleasantly surprised tbh. I am also completely prepared to scale the server up if I get more users on my instance.
Edit: just a follow up, looks like I can scale my instance to a maximum two ways,
“cpu optimized” up to 48 vCPU and 96gb of ram
“Memory optimized” up to 32 vCPU and 256gb of ram
I’m a long way off of the max though now, my server is only 2 vCPU and 4gb memory for now
I’m running a lemmy instance and using about 700mb, up from 500mb before I had any users (though I have maybe a dozen active users lmao)
But I’m not using much CPU at all though. 5% average on a 2core VPS VM. 4 gigs as well. I can scale up a bit and still afford it personally. After that Ill have to ask for donations, and if not enough stop registration.
The scaling is from my cloud provider, hopefully I won’t have to scale up to the max (looks like it’d be like $1300/mo)
700mb of ram isn’t bad at all. Yeah I’m using like 30% cpu on a 2 core right now. Kbin definitely uses more resources than lemmy but I think it has a lot more going on in the tech stack
deleted by creator
Yeah if my instance got busy enough that would be something to definitely consider. I’m using DigitalOcean over something like AWS as well. I wouldn’t want to run it out of my home off of my own network though which is why VPS providers are nice (plus I have all my DNS rules set up through it)
deleted by creator
Bro, colo is cool af. Thank you.
Have any recommendations how to locate one near me? I’ve never heard of something like that before
I’ve never used a colo before; my rack servers live happily in my basement. If you google say “1U network colo <city name>” you’ll find many providers.
Looking at the recent Docker Compose commits, kbin should scale horizontally until it hits limits of postgres.
Its a really good candidate for kubernetes, if you deploy on AWS/Azure and use AKS/EKS with Azure Database/RDS you will be able to flexibly scale far beyond those limits.
I have been meaning to learn Helm for ages. This seems a good excuse.
Unfortunately I thought I’d be brilliant and run the bare metal version of kbin.