K3s on Packet

This is a Terraform project for deploying K3s on Packet.

New projects can build on this Packet K3s Terraform Registry module with:

terraform init --from-module=packet/k3s/packet packet-k3s

This project configures your cluster with:

MetalLB using Packet elastic IPs.
Packet CSI storage driver.

on ARM devices.

This is intended to allow you to quickly spin-up and down K3s clusters in edge locations.

This repository is Experimental meaning that it's based on untested ideas or techniques and not yet established or finalized or involves a radically new and innovative style! This means that support is best effort (at best!) and we strongly encourage you to NOT use this in production.

Requirements

The only required variables are auth_token (your Packet API key), your Packet project_id, facility, and count (number of ARM nodes in the cluster, not counting the controller, which is always set to 1--if you wish to only run the controller, and its local node, set this value to 0).

In addition to Terraform, your client machine (where Terraform will be run from) will need curl, and jq available in order for all of the automation to run as expected.

You will need an SSH key associated with this project, or your account. Add the identity path to ssh_private_key--this will only be used locally to assist Terraform in completing cluster bootstrapping (needed to retrieve the cluster node-token from the controller node).

BGP will need to be enabled for your project.

Clusters

Generating a Cluster Template

To ensure all your regions have standardized deployments, in your Terraform variables (TF_VAR_varname or in terraform.tfvars), ensure that you have set count (number of nodes per cluster), plan_primary, and plan_node. This will apply to all clusters managed by this project.

To add new clusters to a cluster pool, add the new facility to the facilities map:

variable "facilities" {
  type = "map"

  default = {
    newark  = "ewr1"
    narita  = "nrt1"
    sanjose = "sjc1"
  }
}

by adding a line such as:

...
	chicago = "ord1"
   }
}

Manually defining a Cluster, or adding a new cluster pool

To create a cluster manually, in cluster-inventory.tf (this is ignored by git--your initial cluster setup is in clusters.tf, and is tracked), instantiate a new cluster_pool module:

module "manual_cluster" {
  source = "./modules/cluster_pool"

  cluster_name         = "manual_cluster"
  node_count           = "${var.node_count}"
  plan_primary         = "${var.plan_primary}"
  plan_node            = "${var.plan_node}"
  facilities           = "${var.facilities}"
  primary_facility     = "${var.primary_facility}"
  auth_token           = "${var.auth_token}"
  project_id           = "${var.project_id}"
  ssh_private_key_path = "${var.ssh_private_key_path}"
  anycast_ip           = "${packet_reserved_ip_block.anycast_ip.address}"
}

This creates a single-controller cluster, with count number of agent nodes for each facility in the facilities map.

Demo Project

In example/, there are files to configure and deploy a demo project that, once your request is received, returns the IP of the cluster serving your request to demonstrate the use of Packet's Global IPv4 addresses to distribute traffic globally to your edge cluster deployments.

To run the project, you can run the deploy_demo Ansible project by running the create_inventory.sh script to gather your cluster controller IPs into your inventory for Ansible:

cd example/
sh create_inventory.sh
cd deploy_demo
ansible-playbook -i inventory.yaml main.yml

or manually copy example/deploy_demo/roles/demo/files/traefik.sh to your kubectl client machine and run manually to deploy Traefik and the application.

xinity / packet-k3s