Continuous delivery
Continuous delivery is the practice of keeping software in a state where any change that has passed verification could be released to production at any time.
What is continuous delivery?
Continuous delivery is the practice of keeping software in a state where any change that has passed verification could be released to production at any time.
Plain and technical
Every change is built, tested and made ready to ship automatically, so releasing becomes a decision someone makes rather than a project someone runs.
Continuous delivery requires an automated path from commit to a deployable, verified artifact, with each change independently releasable. It is distinct from continuous deployment, in which every passing change is released automatically without a human decision. Both depend on small batch sizes: the benefit comes from being able to attribute a failure to a specific change.
What it looks like in practice
Nearby vocabulary
DevOps
DevOps is a way of working in which the people who build software share responsibility for running it, with automation that makes small change safe.
GitOps
GitOps is an operating model in which a system desired state is declared in Git and an automated process continuously reconciles the running system to it.
Canary deployment
A canary deployment releases a new version to a small share of production traffic first, increasing that share only if measured health stays within threshold.
Blue/green deployment
A blue/green deployment runs the new version as a parallel environment, verifies it, then switches all traffic at once. Reverting is the same switch back.
How DevOpsArk handles continuous delivery
Articles on this subject
Kubernetes deployment strategies: rolling, blue/green and canary
How rolling updates, blue/green and canary deployments differ, what each actually protects you from, and how to choose per environment rather than per opinion.
DevOps automation: what to automate, in what order
A practical sequence for automating a delivery path, which step gives the most back first, which automation tends to be regretted, and how to tell when a stage is genuinely done.
More definitions
See these concepts in a running system
A 30-minute walkthrough against your own infrastructure rather than a slide about the theory.