Ground Control Segment automated deployment and configuration with ANSIBLE and GIT

摘要

In this paper, we present our method to simplify the deployment of a ground control segment which relies on several dozens of machines divided into several validation platforms. We need to automatically instantiate those platforms from a common and centralized configuration. For this purpose, we will describe how we setup a unified configuration management by an Infrastructure as Code approach. We will expose the benefits of this innovative method. The deployment of a ground control segment is made up of several stages: virtual machine creation, operating system installation, computer security and installation of space oriented software. Usually, this deployment is managed by complex operational documentation and homemade scripts adapted to specific needs. A reinstallation can take days and requires a lot of systems engineers. Consequently, a reinstallation is considered as a major risk and therefore rarely done. Cloud oriented approaches intensively use automation to deploy important infrastructures. Methods and tools from these approaches could be interesting for our use case. Infrastructure as Code is one of these methods. It is a way of managing and provisioning computer data centers through machine-readable definition files. It involves the same types of processes used during development to deploy infrastructures and components. In our case, we use a version control system named GIT to track modifications. To achieve the deployment, we rely on a configuration management software named Ansible. GIT is the most popular decentralized version control system for tracking changes in files and coordinating work inside the team. Ansible is an open source configuration management software recently bought by Redhat that enables us to deploy and manage a large set of interconnected machines (physical or virtual). It is based on the popular protocol Secure Shell (SSH) and the largely deployed interpreted language Python. Moreover, it does not require a specific agent installed on each client machine. The definition files describing the different actions managed by Ansible are written in a way that is close to natural language and easily readable. Their organization simplifies the reusability of code. The deployment is idempotent. It means that it can be applied several times without changing the result beyond the initial application. That allows us, when a modification has to be made on the ground control segment, to modify definition files instead of working on a specific server and to reapply the whole configuration via Ansible.