Kubernetes Cluster

Table of Contents


Introduction

While Kubernetes deployments can be quite difficult and can require lots of experience, we provide here a very simple way to provision Kubernetes cluster on the TFGrid.

Example

An example for deploying a kubernetes cluster could be found here

terraform {
  required_providers {
    grid = {
      source = "threefoldtech/grid"
    }
  }
}

provider "grid" {
}

resource "grid_scheduler" "sched" {
  requests {
    name     = "master_node"
    cru      = 2
    sru      = 512
    mru      = 2048
    distinct = true
    public_ips_count = 1
  }
  requests {
    name     = "worker1_node"
    cru      = 2
    sru      = 512
    mru      = 2048
    distinct = true
  }
  requests {
    name     = "worker2_node"
    cru      = 2
    sru      = 512
    mru      = 2048
    distinct = true
  }
  requests {
    name     = "worker3_node"
    cru      = 2
    sru      = 512
    mru      = 2048
    distinct = true
  }
}

locals {
  solution_type = "Kubernetes"
  name          = "myk8s"
}
resource "grid_network" "net1" {
  solution_type = local.solution_type
  name          = local.name
  nodes         = distinct(values(grid_scheduler.sched.nodes))
  ip_range      = "10.1.0.0/16"
  description   = "newer network"
  add_wg_access = true
}

resource "grid_kubernetes" "k8s1" {
  solution_type = local.solution_type
  name          = local.name
  network_name  = grid_network.net1.name
  token         = "12345678910122"
  ssh_key       = "PUT YOUR SSH KEY HERE"

  master {
    disk_size = 2
    node      = grid_scheduler.sched.nodes["master_node"]
    name      = "mr"
    cpu       = 2
    publicip  = true
    memory    = 2048
  }
  workers {
    disk_size = 2
    node      = grid_scheduler.sched.nodes["worker1_node"]
    name      = "w0"
    cpu       = 2
    memory    = 2048
  }
  workers {
    disk_size = 2
    node      = grid_scheduler.sched.nodes["worker2_node"]
    name      = "w2"
    cpu       = 2
    memory    = 2048
  }
  workers {
    disk_size = 2
    node      = grid_scheduler.sched.nodes["worker3_node"]
    name      = "w3"
    cpu       = 2
    memory    = 2048
  }
}

output "computed_master_public_ip" {
  value = grid_kubernetes.k8s1.master[0].computedip
}

output "wg_config" {
  value = grid_network.net1.access_wg_config
}

Everything looks similar to our first example, the global terraform section, the provider section and the network section.

Grid Kubernetes Resource

resource "grid_kubernetes" "k8s1" {
  solution_type = local.solution_type
  name          = local.name
  network_name  = grid_network.net1.name
  token         = "12345678910122"
  ssh_key       = "PUT YOUR SSH KEY HERE"

  master {
    disk_size = 2
    node      = grid_scheduler.sched.nodes["master_node"]
    name      = "mr"
    cpu       = 2
    publicip  = true
    memory    = 2048
  }
  workers {
    disk_size = 2
    node      = grid_scheduler.sched.nodes["worker1_node"]
    name      = "w0"
    cpu       = 2
    memory    = 2048
  }
  workers {
    disk_size = 2
    node      = grid_scheduler.sched.nodes["worker2_node"]
    name      = "w2"
    cpu       = 2
    memory    = 2048
  }
  workers {
    disk_size = 2
    node      = grid_scheduler.sched.nodes["worker3_node"]
    name      = "w3"
    cpu       = 2
    memory    = 2048
  }
}

It requires

  • Network name that would contain the cluster
  • A cluster token to work as a key for other nodes to join the cluster
  • SSH key to access the cluster VMs.

Then, we describe the the master and worker nodes in terms of:

  • name within the deployment
  • disk size
  • node to deploy it on
  • cpu
  • memory
  • whether or not this node needs a public ip

Kubernetes Outputs

output "master_public_ip" {
    value = grid_kubernetes.k8s1.master[0].computedip
}

output "wg_config" {
    value = grid_network.net1.access_wg_config
}

We will be mainly interested in the master node public ip computed IP and the wireguard configurations

More Info

A complete list of k8s resource parameters can be found here.

Demo Video

Here is a video showing how to deploying k8s with Terraform.

Last change: 2024-02-27