RKE deployment on Openstack using Terraform

This CLI allows you to install a Rancher Kubernetes Engine cluster of machines created with Terraform on Openstack. It provides automatic provisioning of Cinder volumes to Kubernetes pods and nginx as LoadBalancer.

Prerequisites

On your machine you need the following requirements:

• Docker
• Python 3.7+
• Set up the environment by sourcing the OpenStack RC file for your project

Installing the CLI

You need python3 installed. I recommend that you use virtualenv for installation of the CLI:

virtualenv venv
source venv/bin/activate

In order to install the CLI please run:

pip install -r requirements.txt
pip install .

You can now explore the different functions provided by the CLI by executing:

rega --help

Deployment configuration

The tool can use different configurations depending on your needs. The default configuration lives in the https://github.com/NBISweden/rega-templates repository so see that for more detailed information.

To create a new default deployment project you can run:

rega init <my-project>
cd <my-project>

And then confgiure that installation according to the underlying repos documentation.

To use a custom deployment repository use the -r/--repository option to the init command (possibly in conjunction with the -b/--branch option).

rega init --repository https://github.com/someorg/somerepo.git --branch stable <my-project>
cd <my-project>

Quick start

Firing up the infrastructure

To spawn the infrastructure execute the apply command.

rega apply

Once the deployment is done, you may explore the cluster:

rega kubectl get nodes -o wide

Releasing resources

You can release the resources by running destroy:

rega destroy

Detailed description

The rega tool looks for scripts in the scripts/ subdirectory of the template repo that is instlled. These scripts have to follow the following naming scheme in order for rega to find them and use them correctly (all other shellscripts are ignored):

./scripts/<stage>_<type>_<name>.sh

Where stage is a number, type is any of init, plan, apply, and destroy and name is a descriptive name of the script.

The rega tool then runs those scripts in the order specified by the stage number when the commands init, plan, apply and destroy is used. The expected behavior of scripts in the different stages are as follows:

Stages

The stage specification is for specifying the order in which the modules should run but also to group actions that belong together. For example with a terraform stage you might have scripts to plan, apply and destroy that part of the infrastructure and by naming them all with the same number. For example:

./scripts/03_plan_network.sh
./scripts/03_apply_network.sh
./scripts/03_destroy_network.sh

Types

init

Run last in the rega init phase. This is where different initialisations can take place. The most common here would be to download terraform modules into the ./terraform_modules directory for example.

Stages are run in order.

plan

Runs when the user enters rega plan. This is roughly equivalent to the terraform plan command, so most of these will just execute a terraform plan command or equivalent in whatever provisioner is used.

Stages are run in order.

apply

Runs when the user enters rega apply. This is for making changes to the infrastructure. Recommended things to run in these stages are terraform apply and ansible-playbook for example

Stages are run in order.

destroy

Runs when the user enters rega destroy. This should destroy the infrastructure.

Stages are run in reverse order

Run only a specific part of the available stages

It's possible to specify what stage to run to plan, apply and destroy by adding the name of the module you want to run to the command line. For example the command:

rega apply network

would run all files that match the file pattern ./scripts/[0-9]+_apply_network.sh in order.

Rancher Server

In order to manage the cluster from the Rancher UI, you can install cert-manager and the Rancher server using a Helm chart. After initialising Helm, you need to add the Helm chart repository that contains charts to install Rancher:

rega helm repo add rancher-stable https://releases.rancher.com/server-charts/stable

Rancher relies on cert-manager version v0.5.2 from the official charts repository. To install it use:

rega helm install cert-manager stable/cert-manager \
  --namespace kube-system \
  --version v0.5.2

And initialise the Rancher server by:

rega helm install rancher rancher-stable/rancher \
  --namespace cattle-system \
  --set hostname=ega.dash.<edge-ip>.nip.io

You can then visit the dashboard in the following address: https://ega.dash.<edge-ip>.nip.io