Skip to content
This repository has been archived by the owner on Oct 5, 2023. It is now read-only.

Latest commit

 

History

History
117 lines (101 loc) · 6.56 KB

README.md

File metadata and controls

117 lines (101 loc) · 6.56 KB

Condor

Terraform Module Registry Docs

Table of Contents

Requirements

  • A funded Vultr account and API Key, should be configured as an environment variable to be consumed by the Terraform provider:
export VULTR_API_KEY="<api-key-here>"
  • The k0sctl executable must be installed and in your executable path.
  • A configured ssh-agent, i.e.:
ssh-add ~/.ssh/id_rsa

Cloud Addons

Quick Start

Usage and input details can be found in the Terraform Module Registry Docs, or use the quickstart below.

  1. Create a main.tf file with the Required Inputs:
# main.tf

module "condor" {
  source                 = "vultr/condor/vultr"
  version                = "2.0.0"
  provisioner_public_key = chomp(file("~/.ssh/id_rsa.pub"))
  cluster_vultr_api_key  = var.cluster_vultr_api_key
  control_plane_firewall_rules = [
    {
      port    = 6443
      ip_type = "v4"
      source  = "0.0.0.0/0"
    }
  ]
}
  • The Control Plane Firewall rule in this example exposes the Kubernetes API globally, it is recommended you configure a more restrictive rule or rules on production clusters.
  • Passing the Cluster API Key as plain text is not recommended for anything beyond testing, use an environment variable as described here.
  1. Configure any Optional Inputs if you wish to change from the defaults.

  2. Deploy

$ terraform init && terraform apply
  1. The Admin Kubeconfig is written locally to your working directory by default as admin-<terraform-workspace>.conf. This may not be desireable in some scenarios, e.g. running Terraform in automation(CI/CD pipelines). If you wish to change this set the write_kubeconfig variable to false. If you wish to retreive the kubeconfig manually, you can run the following after your cluster has been deployed:
$ k0sctl kubeconfig > /path/to/admin.conf
  1. Verify cluster functionality
kubectl --kubeconfig admin.conf get no 
kubectl --kubeconfig admin.conf get po -n kube-system

Firewall Configuration

Control Plane HA VLB Firewall

The Control Plane LB Firewall is configured to allow only what is needed by the cluster as described in the K0s Networking Docs by default. The Kubernetes API will not be accessible without configuring an additional rule or rules(as shown in the quickstart example) via the control_plane_firewall_rules input variable. E.g.:

  control_plane_firewall_rules = [
    {
      port    = 6443
      ip_type = "v4"
      source  = "0.0.0.0/0"
    }
  ]

As also stated in the quickstart, this example rule exposes the Kubernetes API globally, your rules should be more restrictive for production clusters.

Cluster Nodes Vultr Firewall

The cluster nodes(control plane and workers) Vultr Firewall defaults to allowing only SSH globally. This is generally acceptable, however if you would like to restrict access further you may disable this rule by setting the allow_ssh input variable to false then configuring the desired rule/rules outside of this module using the cluster_firewall_group_id output in your rules.

Extensions

Helm

Helm Repositories and Charts may be configured/deployed during initial cluster init via the helm_repositories and helm_charts variables. Note, namespace in a chart definition is required for K0s. Example:

  helm_repositories = [
    {
      name = "argo"
      url  = "https://argoproj.github.io/argo-helm"
    }
  ]
  helm_charts = [
    {
      name      = "argocd"
      chartname = "argo/argo-cd"
      version   = "3.6.8"
      namespace = "argocd"
    }
  ]

Please see the Helm Chart Deployer docs for a comprehensive list of field/parameter values and further details. Note, this feature entails Limitations.

Manifest Bundles

You may deploy any Kubernetes manifests automatically with the K0s Manifest Deployer by placing your manifests in the /var/lib/k0s/manifests directory. Doing so via this module is not supported, however you may use the resulting controller_ips module output as arguments to a separate module that copies your manifests to the specified directory(or as stated in the linked K0s docs, a "stack" subdirectory).

Limitations

  • Shrinking of the Control Plane is not supported, only growing. You will need to manually run k0s etcd leave on all Control Plane nodes with index > 0 prior to shrinking the controller_count. An initial attempt was made to implement this in a destroy time provisioner, however it caused issues when running terraform destroy to destroy the entire plan. This may be revisited at a later date.
  • Etcd running on the Control Plane currently goes over the public network due to k0sctl configuration limitaitons - each controllers respective spec.k0s.spec.storage.peerAddress parameters would require different values. The Cluster Vultr Firewall allows only Control plane nodes to access the Etcd port(2380). Will likely revisit this or look into a PR to K0sctl so that the private network may be used.
  • Helm Charts/Repositories are not removed from your cluster if removed from your Terraform configuration. The manifests must be manually removed from each controller in the /var/lib/k0s/manifests/helm/ directory, and the resources/namespace manually deleted. This makes it less than ideal for continued deployments to your cluster, but great for bootstrapping core applications. A tool such as ArgoCD as used in the Helm example is recommended for long term deployments to your cluster.