Deployment of Kubernetes + shared file system

This repo allows you to deploy Kubernetes plus a shared file system (GlusterFS) using Kubespray v2.3.0 (plus two minor patches).

Requirements

• docker (Manage Docker as a non-root user)
• ability to generate ssh private and public keys
• openstack RC API access file, from Horizon
• a written config for the deployment. See the included config.sh.sample

How to use it

Assemble a folder in your machine with the following files:

• keys: a folder named keys containing key.pub and key.pem (for the private key). If you don't have a set of keys that you want to use, you can create keys executing ssh-keygen -t rsa on linux/macOS Terminal, and rename keys adequately. Please note! Do not use keys that have been generated using a passphrase.
• openstack.rc: OpenStack API access RC file, which you can obtain from your OpenStack Horizon dashboard, under Compute -> API Access. Prefer the v3 file if available.
• config.sh: Bash file with configuration for the deployment (flavours for instances, number of master, etc, see config.sh.sample)

If not available, two folders will be created:

• artifacts: where relevant files are left to access to the cluster, for instance the kubeconfig file for the new cluster, for direct queries from kubectl.
• deployments: terraform state for deployments and other ansible configs.

Then export an environment variable named $PROJECT_DIR pointing to this directory:

export PROJECT_DIR=/full/path/to/my/deployment/project

Keep this directory safe, as it will keep the state of your deployed cluster, config file and keys, which are relevant for accessing it, modifying it or destroying it.

Finally, to run it:

$ bash start-env-for-running.sh

It might need to pull the first time, which might make you miss the following instruction:

Then inside the container, execute ./run-deployment.sh

After some 20 minutes it should provide you with access instructions for both kubectl and ssh.

Access the cluster

Inside the same container

First, make sure that the $KUBECONFIG is set to artifacts/admin.conf, then execute:

$ kubectl <whatever-command-you-want>

Outside the container

Set $KUBECONFIG to $PROJECT_DIR/artifacts/admin.conf, then:

$ kubectl <whatever-command-you-want>

ssh-access

See artifacts/ssh-connect file for details.

Destroying the cluster

Execute inside the same container, or after running start-env-for-running.sh in an environment where $PROJECT_DIR is set to the correct directory:

$ bash run-destroy.sh

Hint: if possible, limit all your interaction with the cluster through kubectl, which will make this cluster more easily "disposable".