Skip to content

Latest commit

 

History

History
83 lines (55 loc) · 5.03 KB

README.md

File metadata and controls

83 lines (55 loc) · 5.03 KB
Raw

Terraform - Provision a GKE Cluster with Cloudflare Ingress and ArgoCD

License ShellCheck

This repo contains three Terraform modules to provision a GKE cluster (with VPC and subnet), then deploy Helm charts and Kubernetes manifests.

The included deployments are designed for a fully-functioning Ingress controller that works with Cloudflare.

Note: the GCP module in this repo is a modified fork of learn-terraform-provision-gke-cluster; the MPL 2.0 license is adhered to.


What is deployed?

What is needed to deploy?

The deploy.sh script does everything needed in order to deploy the full environment in the correct order. In order to use this script, you need a Unix shell (the script is written for POSIX shell - designed to have wide compatibility in Linux, macOS and WSL environments).

In addition, you will need to have the following packages installed:

For Cloudflare, you should own a domain and register it in Cloudflare. Then generate an API token with the permissions described here: https://cert-manager.io/docs/configuration/acme/dns01/cloudflare/#api-tokens

How to deploy this?

  1. In your shell, run gcloud init

  2. Enable billing for the GCP project you're using: see Google Cloud documentation on enabling billing for a project. GCP free trial can be used (if eligible).

  3. Edit the terraform.tfvars.example file:

  • Use the command: cp "terraform.tfvars.example" "terraform.tfvars"
  • Edit terraform.tfvars using a text editor. Make sure to add an appropriate GCP project ID (according to what you configured in step 1).

  1. Run the deploy.sh script

  2. When running the script, the GKE cluster would be provisioned first, together with a sealed-secrets controller. You would then be asked to provide your Cloudflare API token which will be configured as a sealed secret and used for the DNS-01 challange and ExternalDNS.

  3. When the script completes, run kubectl get pods -A to ensure everything deployed correctly.

How to destroy this?

Destroying is much simpler. Just run terraform destroy from the project directory.


FAQ

Can I use this without Cloudflare?

Yes, but you will need to modify the cluster issuers and ExternalDNS to work with a different DNS provider.

Can I use this without GCP?

In theory yes but it will require the modules to be heavily rewritten.

Can I use this in production?

This is not recommended. These modules are designed only for learning (e.g. using a GCP free trial).

Can I deploy other Helm charts?

Yes, in the "helm_releases" folder, add any additional Helm chart as a tf file containing a "helm_release" resource.

Can I deploy other Kubernetes manifests?

Yes, in the "kubernetes_manifests" folder, add any additional Kubernetes manifest as a tf file containing a "kubernetes_manifest" resource. Make sure the manifest is formatted as JSON, not YAML.

How do I access the Ingress?

Apply an approriate Ingress resource for your service (see Kuberenets documentation) and add the following annotations:

annotations:
    external-dns.alpha.kubernetes.io/hostname: "your.domain,*.your.domain"  # MODIFY THIS
    cert-manager.io/cluster-issuer: letsencrypt-prod                        # or letsencrypt-staging

The first annotation updates the DNS records using ExternalDNS. The second annotation uses cert-manager to provision a Let's Encrypt certificate (use either prod or staging).

I am getting an SSL/TLS error or redirect error while accessing my domain after applying the ingress resource. How can I solve this?

See Cloudflare Docs: ERR_TOO_MANY_REDIRECTS

It is recommended to set the SSL/TLS encryption mode in Cloudflare to Full or Full (strict); if using a staging or self-signed certificate, use Full. With a prod certificate, both modes can be used. If not using any certificate, use Cloudflare's "Flexible" mode.