- Continuous Delivery: Reliable Software Releases through Build, Test, and Deployment Automation by Jez Humble, David Farley
- Infrastructure as Code: Dynamic Systems for the Cloud Age by Kief Morris
- Continuous Delivery by Martin Fowler
Continuous delivery is software development practice, in general, and an agile software development practice, in particular, which encourages building software in such a way that it can be released in production at any time. Because of the safety mesh provided by CD, the system is released in production often. Before releasing in production, the system is deployed in an environment similar to production and tested there. As such, Continuous Delivery extends the unit testing performed by Continuous Integration, ensuring full production readiness through integration testing. The essence of CD is to maximize the scope of testing initiated by CI, by exercising the entire system in its natural habitat. If properly conducted, an integration defect is caught immediately, instead of waiting for it to materialize later in production. Only the code that deploys and runs correctly in one or more test environments is promoted to production.
CD pipelines deploy the artifacts produced by the CI pipelines. Continuous Delivery is some times referred to as Continuous Deployment.
CD best practices:
- version code and configuration
- version environment
- build binaries once
- automate everthing
- smoke test deployments
- deploy to all environments the same way
- create disposable environments
A core principle of CD is to never change the code after the build stage.
Amazon take on Continuous Delivery: https://aws.amazon.com/devops/continuous-delivery/
Continuous Delivery vs. Continuous Deployment
Within the context of these articles, continuous deployment covers the part of the software delivery process that consists of placing - deploying - artifacts and configuration into a target environment, possibly production, every time there is an artifact or a configuration change, or at any moment an operator may choose so. The artifacts are the products of a continuous integration process. Since the artifacts are continuously built and published as result of upstream changes, automated deployments are triggered as result, by default. However, not every deployment is a production deployment, and the system must allow for selectivity and flexibility in deciding where and how these artifacts are deployed. A specific feature may be continuously deployed to a Stage environment while it is being worked on and tested in conditions similar to production, and thus iteratively refined, without actually being released to production users. Multiple continuous integration and continuous deployment cycles can be exercised as part of such an iterative deployment process, but the change may end not being delivered to production users. The system must allow flexibility and manual intervention for this kind of decisions.
Continuous delivery is a more generic concept. Aside from its intrinsic technical implications, it also has business implications. Continuous delivery is about propagating a business-level change all the way to production and in the hands of users. Such change translates internally in changes of code, data structures, configuration, etc. and those changes trigger continuous integration and continuous deployment cycles. Continuous delivery requires the capability to do continuous integration and deployment. Implementing continuous delivery means making sure that software is always production ready throughout its entire lifecycle and any build could potentially be released to production users at the touch of a button and using a fully automated process in a matter of seconds or minutes. This in turn relies on comprehensive automation of the build, test suite and deployment infrastructure. In the world of continuous delivery, a developer is only done when the feature is working in production and produces value for users.
More thoughts on this subject:
A delivery repository exposes packaged artifacts or code for use with the infrastructure tools that create infrastructure in the target environment or that deploy the applications in the application runtime. The delivery repository usually stores multiple version of a given project's code. The delivery repositories could be general purpose storage solutions such as file servers, S3 buckets, tool-specific repositories, such as Terraform Registry or specialized artifact repositories like Nexus and Artifactory or ORAS.
Especially for infrastructure, given that the infrastructure code is already stored in the source repository, and many infrastructure code tools don't have a package format and a toolchain for threading their code as a release, many teams apply code to environments directly from the source repository. See:
- Hosted Services
- Cloud Platform Services
- AWS CodePipeline
- Azure Pipelines
Fan-in Pipeline Design
TO PROCESS: Modeling Deployment Pipelines: Build Propagation using Fan-in/Fan-out https://www.gocd.org/2017/04/17/build-propagation-using-fan-in-fan-out.html
Used in Delivery-time integration.