Skip to content

Latest commit

 

History

History
66 lines (39 loc) · 4.61 KB

cluster-auto.md

File metadata and controls

66 lines (39 loc) · 4.61 KB
Raw

[AUTOMATIC] Deployment steps

Every command must be run at the root of the repository

Prerequisites

Docker and docker-compose must be installed on your computer.

Create Terraform's state

First, we need a s3 bucket to store the Terraform's state, so that it can be available everywhere (and not only on your computer). If you already have a bucket, you can skip this step.

This repository provides a Terraform to create a bucket on OVH, but not on Scaleway. For this step, you will need OVH API credentials (application key, secret key and consumer key, as well as the project id in which you will create the bucket, see here to generate a token). You must add the following rights and replace {serviceName} by your OVH's Public Cloud project id :

  • GET /cloud/project/{serviceName}/*
  • PUT /cloud/project/{serviceName}/*
  • POST /cloud/project/{serviceName}/*
  • DELETE /cloud/project/{serviceName}/*

Then execute the corresponding script : bin/init-bucket.sh, after entering all the required information, it will create a bucket on OVH;

And finally, save the provided credentials access_key, secret_key and bucket_name, you will need them for the next step.

Create and provision the cluster

Configure the backend

Now we've got our s3 bucket, we have to setup Terraform's backend, where it stores its state. For this, we will use the s3 bucket we just created (or the one you already have).

Run bin/bootstrap-backend.sh <your provider> to create the backend. It will create a backend.conf file, which will be used by Terraform to store its state in the s3 bucket. Replace <your provider> either with ovh or scaleway.

If you used the previous script to generate the bucket, here are some information you need:

  • Region: gra
  • Endpoint: https://s3.gra.io.cloud.ovh.net/
  • Skip region validation: true
  • Skip credentials validation: true

Provide the correct information

Terraform needs a few variables to create your cluster, please run bin/bootstrap.sh <your-provider> and provide the desired values for each parameter. You will need:

  • The hostname for several services: ArgoCD, Grafana, Vault (if installed)
  • A already existing S3 bucket for Velero (you can use the state_bucket terraform script to create a S3 bucket for Velero)
  • ArgoCD needs a Git repository with HTTPS credentials for access. You can use a private repository, or a public one. If you use a private repository, you will need to provide the HTTPS credentials (username and password). If you use a public repository, you can leave the username and password empty.
  • API keys for your provider:
    • For OVH, see here
    • For Scaleway, see here

The script will prompt for the most common variables. By default, some variables are not prompted (and their default value is then used). If you wish, you can look into the variables.tf and the variables-common.tf files to see all the variables that can be set. Simply add them to the terraform.tfvars file.

Deploy the cluster

After your terraform.tfvars file has been successfully created, you can now deploy the cluster. Run bin/terraform-init.sh <your provider> to initialize Terraform. After this, run bin/terraform-plan.sh <your provider>, the output shows you what Terraform will do. If you are satisfied with the plan, run bin/terraform-apply.sh <your provider> to deploy the cluster. (Please ignore the output of the command beginning with 'To perform exactly these actions...')

While running, the terraform-apply.sh script may crash (especially with OVH). If so, analyze the error. If it is related to timeouts or server errors, simply re-run the script. (if you encounter errors re-running terraform-apply.sh, try running terraform-plan.sh before). The script may last more than 10 minutes, please be patient.

Warning: If the script were to crash, make sure Terraform has not been creating ressources (e.g. a k8s cluster) in the background (which has not been linked to the state due to the crash). If so, you will have to delete them manually.

At the end of the script, please make the needed changes on your DNS (adding the ingress IP to the needed domains), you may be able to retrieve your Kubeconfig file with the following command: bin/get-kube-config.sh <your provider>.

Destroy the cluster

With: bin/terraform-destroy <your provider>. Warning: there is no confirmation, it will destroy the cluster immediately.

Next step → Configure Hashicorp Vault