CI/CD
Every commit becomes a tested image, then a release
Continuous integration builds and tests. Continuous delivery keeps it ready to release.
Application delivery is the road from a developer's change to a running release. It starts with a commit, pushed to Git. The pipeline builds it into an image and tests it. Doing that for every commit is continuous integration. The image is then published to a registry and rolled out to the cluster. Keeping every tested change ready to release is continuous delivery. When even the release needs no person, it is called continuous deployment.
CI/CD · Going deeper
The three terms and the tools
Delivery keeps a manual approval. Deployment does not.
GitOps
The cluster pulls its desired state from Git
Git holds the desired state. An agent in the cluster pulls it and corrects drift.
How does the last step, deploy, actually happen? The older way is push. The pipeline holds the keys to the cluster and applies the change. If someone later changes the cluster by hand, nothing notices. GitOps turns it around. An agent inside the cluster, Argo C D or Flux, watches a Git repository and pulls the desired state in. And it keeps comparing. When the cluster drifts from Git, the agent puts it back.
GitOps · Going deeper
The GitOps principles, Argo CD and Flux
Declarative, versioned, pulled, reconciled.
Check yourself
What does the agent do next?
Packaging
Helm fills in templates. Kustomize patches manifests
Helm: chart plus values gives a release. Kustomize: base plus overlay gives manifests.
The same application runs in test and in production with different settings. Two tools handle that. Helm is a package manager. A chart holds templates with blanks. You supply values, Helm fills them in, and the result is a release that you can upgrade and roll back. Kustomize uses no templates. You keep plain manifests as a base, and a small overlay of patches for each environment. Kubectl merges them when you apply with dash K.
Packaging · Going deeper
Helm and Kustomize terms
Chart, release, repository; base and overlay.
Release strategies
Rolling, blue-green and canary move traffic differently
Rolling replaces step by step. Blue-green switches at once. Canary starts small.
Now, how users are moved to a new version. Each bar shows the share of traffic on the new version, as time passes. A rolling update replaces pods a few at a time, so the share climbs in steps. Blue green starts a complete second copy, which receives nothing at first. Then all traffic switches in one move. A canary sends a small share first, here five percent. If the numbers look healthy, it sends more, and then everything.
Release strategies · Going deeper
Strategies compared
Capacity and rollback differ. Argo Rollouts and Flagger automate them.
Check yourself
Which release strategy is this?
Debugging a rollout
A bad release stalls, and one command undoes it
Each revision is a ReplicaSet. kubectl rollout undo returns to the last one.
Last, what happens when a release goes wrong. A deployment named web runs three healthy pods, through a replica set. That is revision one. Someone releases a new image with a wrong tag. A second replica set appears, and its first pod cannot pull the image. The rollout stalls. But the old pods are still serving, so users notice nothing. One command, rollout undo, returns to revision one. The old replica set was kept for exactly this.
Debugging a rollout · Going deeper · 1 of 2
Rollout and debugging commands
rollout status, history, undo, and kubectl debug.
Debugging a rollout · Going deeper · 2 of 2
Clue words for delivery and debugging
GitOps, Helm, canary and rollout undo.
Check yourself
Which command fits what the team needs?
Lesson complete
You finished “Application delivery and debugging”
Answer the questions along the way to see your score here.