Thank you for your reply @metadata, yes, we have this machine and going to install jitsi kubernetes so that we can scale horizontally in the future.
Is Microk8s good option for us? or which is best?
Can you tell me how much we should pay monthly for kubernentes?
If you are new to K8s then start with services like EKS, IKS, etc since running self-managed k8s requires in-depth knowledge of k8s.
Start with less configuration. running this huge machine is going to cost you a lot of money. Start with 8Gb RAM and 8vCPU per worker node. Go with 5 worker nodes and then scale these worker nodes dynamically on the basis of load.
The first challenge in running self-managed K8s is to make them highly available(min 3 master nodes).
OR
if you still want to go ahead with the same machine then use k3s because it is lightweight and easily configurable and manageable. use tools like K3sup to install K3s on your machine.
K3s is far much better than Microk8s. It isn’t easy to make MicroK8s highly available. Sharing my experience here.
I’m using IKS(IBM classic K8s) with 5 worker nodes (8vCPU and 8Gb Memory, 1Gbps Bandwidth) and It cost me approx $1.2K per month(because of 1Gbps Bandwidth). I’m using it most of the time for my testing only. The benefit of using this kind of service is that the only thing you have to worry about is your workload everything else is taken care of by service providers.
Is 1.2K(your cost) for only Kubernetes or Kubernetes+5 nodes?
We’re going to planning extra cost from kubernetes installation.
So can you provide me K3s cost? (without including server cost)
Do we have to create multiple nodes in this server?
Or is node a cloud machine?
Do we have to purchase more nodes if we can’t handle many users with current server?
If we only use current one server, then don’t we need to install kubernetes?
PS: This is already running web server handling big concurrent users , so we’re going to install jitsi on same server.
K8s + 5 worker nodes. Actually for services like EKS, IKS you will get only worker nodes. master nodes will be managed by service providers so that you can focus only on your deployment.
K3s is an Open Source product by SUSE(Rancher Labs).
if talking about self-managed then yes. Because you must run any k8s cluster with minimum 3 Master nodes(if using it in production)
A worker node is a Virtual Machine.
Cloud providers will spin up the nodes for you based on the load but yes it is going to cost you some money.
if you want to deploy jitsi on k8s then you need to install k8s on top of it.
@metadata, Should I install k3s in bare-metal server filesystem or in docker file system?
And is k3s good enough to serve huge people?
seem to be used for only development purpose.