Matt McEuen 8387edcf0e Add some clarifications to AIO dev guide

Add some clarifications to the AIO dev guide, based on real-world
points of confusion that have arisen:
- Do I *have* to use Ubuntu 16.04?
- What docker should I use?
- Should I run the scripts as root?

Change-Id: I8affdde16b63cc078aa496bf168154c93c39b3bf

2017-12-30 08:21:33 -06:00

11 KiB

Raw Blame History

All-in-One

Overview

Below are some instructions and suggestions to help you get started with a Kubeadm All-in-One environment on Ubuntu 16.04. Other supported versions of Linux can also be used, with the appropriate changes to package installation.

Requirements

System Requirements

The recommended minimum system requirements for a full deployment are:

16GB of RAM
8 Cores
48GB HDD

For a deployment without cinder and horizon the system requirements are:

8GB of RAM
4 Cores
48GB HDD

This guide covers the minimum number of requirements to get started.

All commands below should be run as a normal user, not as root. Appropriate versions of Docker, Kubernetes, and Helm will be installed by the playbooks used below, so there's no need to install them ahead of time.

Warning

By default the Calico CNI will use 192.168.0.0/16 and Kubernetes services will use 10.96.0.0/16 as the CIDR for services. Check that these CIDRs are not in use on the development node before proceeding, or adjust as required.

Host Configuration

OpenStack-Helm uses the hosts networking namespace for many pods including, Ceph, Neutron and Nova components. For this, to function, as expected pods need to be able to resolve DNS requests correctly. Ubuntu Desktop and some other distributions make use of mdns4_minimal which does not operate as Kubernetes expects with its default TLD of .local. To operate at expected either change the hosts line in the /etc/nsswitch.conf, or confirm that it matches:

hosts:          files dns

Packages

Install the latest versions of Git, CA Certs & Make if necessary

../../../../tools/deployment/developer/00-install-packages.sh

Clone the OpenStack-Helm Repos

Once the host has been configured the repos containing the OpenStack-Helm charts should be cloned:

#!/bin/bash
set -xe

git clone https://git.openstack.org/openstack/openstack-helm-infra.git
git clone https://git.openstack.org/openstack/openstack-helm.git

Deploy Kubernetes & Helm

You may now deploy kubernetes, and helm onto your machine, first move into the openstack-helm directory and then run the following:

../../../../tools/deployment/developer/01-deploy-k8s.sh

This command will deploy a single node KubeADM administered cluster. This will use the parameters in ${OSH_INFRA_PATH}/tools/gate/playbooks/vars.yaml to control the deployment, which can be over-ridden by adding entries to ${OSH_INFRA_PATH}/tools/gate/devel/local-vars.yaml.

Helm Chart Installation

Using the Helm packages previously pushed to the local Helm repository, run the following commands to instruct tiller to create an instance of the given chart. During installation, the helm client will print useful information about resources created, the state of the Helm releases, and whether any additional configuration steps are necessary.

Install OpenStack-Helm

Note

The following commands all assume that they are run from the openstack-helm directory and the repos have been cloned as above.

Setup Clients on the host and assemble the charts

The OpenStack clients and Kubernetes RBAC rules, along with assembly of the charts can be performed by running the following commands:

../../../../tools/deployment/developer/02-setup-client.sh