OpenStack Script for Cross-Cloud Continuous Integration

Getting Started

Prerequisites

Building the CNCF/CICD cross-cloud container requires a recent version of Docker. You will also need credentials for an OpenStack cloud with Keystone, Nova, Neutron. LBaaSv2 is not yet supported in this module.

Building the container

In the top-level directory,

docker build . --tag provisioning

Configuring your cloud

You will need the following credentials for your OpenStack cloud. These credentials will used to provision the Kubernetes cluster, and will also be used for the OpenStack provider to Kubernetes. In your environment, set the following OpenStack environment variables:

• OS_AUTH_URL
• OS_REGION_NAME
• OS_USER_DOMAIN_NAME
• OS_USERNAME
• OS_PROJECT_NAME
• OS_PASSWORD

Then assign them to Terraform variables like this:

• TF_VAR_os_auth_url=$OS_AUTH_URL
• TF_VAR_os_region_name=$OS_REGION_NAME
• TF_VAR_os_user_domain_name=$OS_USER_DOMAIN_NAME
• TF_VAR_os_username=$OS_USERNAME
• TF_VAR_os_project_name=$OS_PROJECT_NAME
• TF_VAR_os_password=$OS_PASSWORD

This deployment assumes you have a keypair available to inject into all of the instances. The default name is K8s, but you may set it with the variable TF_VAR_keypair_name.

This project assumes the existence of recent CoreOS release. It defaults to the image named CorsOS 1520.8.0, but can be configured with the variables:

• TF_VAR_lb_image_name
• TF_VAR_master_image_name
• TF_VAR_worker_image_name.

Similarly flavors (defaulting to v1-standard-1) can be set with:

• TF_VAR_lb_flavor_name
• TF_VAR_master_flavor_name
• TF_VAR_worker_flavor_name

Auto scaling is not supported, but you can control the number of master and worker nodes with the variables:

• TF_VAR_master_node_count
• TF_VAR_worker_node_count

Network parameters include:

• TF_VAR_bastion_floating_ip_pool
• TF_VAR_external_network_id

As this is a multi-master deployment, a single load balancer is created to proxy non-terminated TLS requests to the K8s master nodes. Given the sporadic availability of LBaaSv2/Octavia across OpenStack clouds, the load balancer is created manually rather than relying on the LBaaSv2 API.

Deploying

The following script will deploy Kubernetes into your cloud:

#!/bin/bash
docker run \
  -v $(pwd)/data:/cncf/data \
  -e TF_VAR_os_auth_url=$OS_AUTH_URL \
  -e TF_VAR_os_region_name=$OS_REGION_NAME \
  -e TF_VAR_os_user_domain_name=$OS_USER_DOMAIN_NAME \
  -e TF_VAR_os_username=$OS_USERNAME \
  -e TF_VAR_os_project_name=$OS_PROJECT_NAME \
  -e TF_VAR_os_password=$OS_PASSWORD \
  -e CLOUD=openstack \
  -e COMMAND=deploy \
  -e NAME=cross-cloud \
  -e BACKEND=file \
  -ti provisioning

Connecting with kubectl

The deployment script will drop a kubeconfig file into $(pwd)/data/cross-cloud. As root, make a copy of it into a local location, change the ownership to your local user, and export the KUBECONFIG environment variable. For example:

sudo cp $(pwd)/data/cross-cloud/kubeconfig ~/.
sudo chown $(whoami):$(whoami) ~/kubeconfig
export KUBECONFIG=~/kubeconfig
kubectl get nodes

Destroying

The following script will destroy the Kubernetes deployment:

#!/bin/bash
docker run \
  -v $(pwd)/data:/cncf/data \
  -e TF_VAR_os_auth_url=$OS_AUTH_URL \
  -e TF_VAR_os_region_name=$OS_REGION_NAME \
  -e TF_VAR_os_user_domain_name=$OS_USER_DOMAIN_NAME \
  -e TF_VAR_os_username=$OS_USERNAME \
  -e TF_VAR_os_project_name=$OS_PROJECT_NAME \
  -e TF_VAR_os_password=$OS_PASSWORD \
  -e CLOUD=openstack \
  -e COMMAND=destroy \
  -e NAME=cross-cloud \
  -e BACKEND=file \
  -ti provisioning