Continuous Delivery Software Deployment and Configuration Management for Critical Operations Environments

Kurzfassung

In space operations environments, both software development management and system configuration management are essential to a zero-defect ground system. The two disciplines interface at release and deployment management, where source code is built into a software package, which is then deployed onto hardware underlying (ground) system configuration management. Software development workflows are typically well-specified and supported by a number of commercial and open-source solutions, e.g., version control and change management via Git and Jira, or continuous integration via Jenkins. System configuration management, governing which software package should be, and actually is installed on which hardware, can also refer to very good tool support, e.g., Puppet or Chef for configuration deployment management, or the HP uCMDB as configuration management database. These tools control, verify and document the ground system software deployment and configuration, ensuring that the system performs as intended over its projected life cycle. The missing link, both in process specification and in tool support, lies between the software development management system and the ground system configuration management system. Until now, the delivery of released software packages to the actual deployment targets consists of several manual error-prone steps, it has to overcome technical barriers (for example network barriers between staging areas and secure operational environments), and is often not transparent for system engineers. To overcome these barriers and close the missing link with appropriate tool support is the challenge we address in this paper. One approach to the problem is the principle of Continuous Delivery, a set of practices designed to ensure that software releases are delivered quickly and safely to a production environment and into the hands of users for deployment, first on test and later on operational chains. Continuous Delivery has to be distinguished here from Continuous Deployment, in which a release that has passed automated testing is also deployed automatically, which is clearly not desired when operational or customer constraints require the deployment to be triggered manually and precisely timed. Many commercial and open-source solutions accomplish part of the job. Tools such as Jenkins support deployment pipelines, Sonatype Nexus offers a repository for software libraries and build artifacts and JFrog Artifactory is a commercial binary repository. Deployment processes, however, are highly business-specific and the available solutions either suffer from being too general and heavy-weight, or from not being compatible with existing workflows and environments. The solution presented in this paper builds on the existing infrastructure and open-source tool chain, supplemented by an in-house software development. It enables us to reliably access the correct software release builds and deploy them easily on the appropriate targets, thereby also serving as a backup and history repository for the deployment packages. We have described in a previous paper how the process of manual configuration management can be automated to be made less error-prone and more complete by scanning hardware for software and transferring this information to a CM database. This automated approach relies on proper deployment management of the installed software packages, as a scan of the hardware will only return software that has been installed as a package, such as RPM (RPM Package Manager) or MSI (MicroSoft Installer). The automatic building of such packages therefore enables and improves configuration deployment and verification. The tool-enabled workflow makes both configuration management and software development management more efficient and transparent. By closing the gap between continuous integration and deployment in an operational environment, we make both configuration management and software release management more efficient and transparent.