Version control systems like Git, Subversion, and Mercurial provide a logical means to organize files and coordinate their creation, controlled access, updating, and deletion across teams and organizations. Version control is closely related to automation. In fact, automation and continuous integration rely on these files for the source code of the automation itself, as well as the configuration to be automated and the data to be distributed.
In order to improve software delivery, teams need to use version control for source code, test and deployment scripts, infrastructure and application configuration information, and the many libraries and packages they depend upon. In the version control system, teams must be able to query the current (and historical) state of their environments. Version control also offers direct benefits such as disaster recovery and auditability.
Research shows that comprehensive use of version control, among other capabilities, predicts continuous delivery. In particular, version control helps you meet these critical requirements:
Reproducibility. Teams must be able to provision any environment in a fully automated fashion, and know that any new environment reproduced from the same configuration is identical. A prerequisite for achieving this goal is having the scripts and configuration information that are required to provision an environment stored in a shared, accessible system.
Traceability. Teams should be able to pick any environment and determine quickly and precisely the versions of every dependency used to create that environment. They should also be able to compare two versions of an environment and see what has changed between them.
These capabilities give teams several important benefits:
Disaster recovery. When something goes wrong with an environment—for example, a hardware failure or a security breach—teams need to be able to reproduce that environment in a deterministic amount of time in order to be able to restore service.
Auditability. To demonstrate the integrity of the delivery process, teams must be able to show the path backward from every deployment to the elements it came from, including their version. You enable this through comprehensive configuration management combined with deployment pipelines.
Higher quality. The software delivery process is often subject to long delays waiting for development, testing, and production environments to be prepared. When this preparation can be done automatically from version control, teams can get feedback on the impact of their changes more rapidly, enabling teams to build quality into their software.
Capacity management. When teams want to add more capacity to their environments, the ability to create reproductions of existing servers is essential. This capability enables the horizontal scaling of modern cloud-based distributed systems.
Response to defects. When teams discover a critical defect, or a vulnerability in some component of their system, they need to release a new version of their software as quickly as possible. Storing all artifacts in version control means teams can roll back to a previously verified working state quickly and reliably.
As environments become more complex and heterogeneous, it's progressively harder to achieve these goals. It's impossible to achieve perfect reproducibility and traceability for a complex enterprise system (at a minimum, every real system has state). Thus, a key part of configuration management is working to simplify the architecture, environments, and processes to reduce the investment required to achieve the expected benefits.
How to implement version control
When implementing version control, we recommend that you start by defining in measurable terms the goals you want to achieve. This allows you and your teams to determine the best path to reach those goals. This approach also lets you change direction or reassess those goals if the path you choose is too expensive or takes too long.
A version control system records changes to files stored in the system. These files can be source code, assets, or other documents that might be part of a software development project. Teams make changes in groups called commits or revisions. Each revision, along with metadata related to the revision (such as who made the change and when), is stored in the system. This allows teams to commit, compare, merge, and restore to previous revisions. It also minimizes risks by establishing a way to revert objects in production to previous versions.
Teams must be able to restore production services repeatedly and predictably (and, ideally, quickly) even when catastrophic events occur, so they must check in the following assets to their shared version control repository:
- All application code and dependencies (for example, libraries and static content)
- Any script used to create database schemas, application reference data, and so on
- All environment creation tools and artifacts described in the previous step (for example, VMware or AMI image building scripts or Chef recipes)
- Any file used to create and compose containers (for example, Docker files and buildpacks)
- All supporting automated tests and any manual test scripts
- Any script that supports code packaging, deployment, database migration, and environment provisioning
- Supporting project artifacts (for example, requirements documentation, deployment procedures, and release notes)
- Container orchestration (for example, Kubernetes configuration, Mesos configuration, and Docker Swarm configuration)
- All cloud configuration files (for example, AWS Cloudformation templates, Cloud Deployment Manager configuration, Microsoft Azure Stack DSC files, OpenStack HEAT, Terraform files, and Pulumi stacks)
- Any other script or configuration information required to create infrastructure that supports multiple services (for example, enterprise service buses, database management systems, DNS zone files, configuration rules for firewalls, and other networking devices)
Version control can take many forms, apart from traditional file-based version control systems like Git. Teams might have multiple repositories for different types of objects and services that are versioned, labeled, and tagged alongside their source code. For instance, they might store large virtual machine images, ISO files, compiled binaries, and so forth in artifact repositories such as Nexus or Artifactory. Alternatively, they might put objects in blob stores such as Cloud Storage or Amazon S3, or they might put Docker images into Docker registries. These approaches meet the requirements of reproducibility and traceability, and provide the same benefits.
More than re-creating any previous state of the production environment, teams must also be able to re-create the preproduction and build processes. Consequently, they also need to check into version control everything their build processes rely on, including tools and the environments they depend upon.
Common pitfalls in version control
The most common pitfall in using version control is limited application or use; in other words, applying version control only to software application code. Best practice requires the ability to reproduce all testing and production environments, including the software deployed on them, in a fully automated fashion by using scripts, source code, and configuration information that's stored in version control.
Ways to improve version control
You can improve version control in many ways. Here a few we recommend:
- Ensure that every commit to version control triggers the automated creation of packages that can be deployed to any environment using only information in version control.
- Make it possible to create production-like test environments on demand using only scripts and configuration information from version control, and to create packages using the automated process described in the previous approach.
- Script testing and production infrastructure so that teams can add capacity or recover from disasters in a fully automated fashion.
As you implement a version control system, focus on your constraints. For example, what's the biggest blocker to the fast flow of changes from version control to production? Are your builds too slow? Is it hard to re-create deployable packages? Is it difficult to create production-like test environments? These constraints can make it hard to achieve your goals, and might indicate a problem to work on with your system's architecture.
Ways to measure version control
To measure how effectively your teams are using version control in their systems, try these recommendations:
Application code. Do you use version control for application code? What percentage of application code do you store in version control? How easily and quickly can a team recover application code from the version control system?
System configurations. Do you use version control for system configurations? What percentage of system configurations do you store in version control? How easily and quickly can teams reconfigure systems from version control?
Application configuration. Do you use version control for application configurations? What percentage of application configurations do you store in version control? How easily and quickly can teams reconfigure applications from code in the version control system?
Scripts for automating build and configuration. Do you keep scripts for automating build and configuration in version control? What percentage do you store in version control? How quickly and easily can you reprovision systems by using scripts from version control?
These recommendations are just the beginning, but they're essential, so we suggest that you start with them and learn how to do them well. Then review this article and identify additional artifacts that you use in developing and delivering software, and ask similar questions: What percentage of those artifacts are in version control? How quickly and easily can your team deploy new systems or configurations using assets from version control?
- For links to other articles and resources, see the DevOps page.