Kubernetes rolling deployment flow

Kubernetes deployment flow diagram

A worked example to adapt. Rename steps, add branches and owners to match how your team actually works.

flowchart TD
  s(["New image pushed to registry"])
  man["Update image tag in manifest"]
  app["Apply deployment"]
  pod["Start new pods"]
  d1{"Readiness probe passes?"}
  svc["Service sends traffic to new pods"]
  old["Terminate old pods"]
  d2{"All replicas updated?"}
  st["Rollout stalls: new pod gets no traffic"]
  rb["On-call or tooling rolls back to previous revision"]
  prev(["Previous version serving"])
  e(["Rollout complete"])
  s --> man
  man --> app
  app --> pod
  pod --> d1
  d1 -->|Yes| svc
  d1 -->|No| st
  st --> rb
  rb --> prev
  svc --> old
  old --> d2
  d2 -->|No| pod
  d2 -->|Yes| e

About this kubernetes deployment flow diagram

A rolling update replaces pods a few at a time, so the application stays available. The key control is the readiness probe: new pods only receive traffic once they report ready. This flowchart shows that loop, and what happens when a new pod never becomes ready: the rollout stalls and is rolled back explicitly.

It describes a standard Deployment with a rolling update strategy. Blue-green and canary releases need extra steps.

Step by step

1. Update. Point the Deployment at the new image tag and apply it.
2. Start pods. New pods start alongside old ones within the surge limits.
3. Check readiness. Pods that pass the readiness probe join the Service. A pod that never becomes ready gets no traffic and the rollout stalls until someone or some tooling rolls it back.
4. Replace. Old pods are terminated as new ones become ready.
5. Repeat. The loop continues until every replica runs the new version.

How to make it in flow-chart.io

1. Start from the example. Edit the text in the generator box above so it names your own cluster and deployment, then press Generate. The free preview needs no sign-up.
2. Add the branches. Add the decision points that matter for you, for example canary release? or database migration first?. Each decision becomes a diamond with labeled outcomes.
3. Assign owners. Save the diagram to the editor (free account) and rename steps to show who does what: CI system, cluster and on-call engineer.
4. Share or export. Share a read-only view link: people with the link can view the diagram but not edit it. Downloads as PNG, SVG or PDF are part of Flow Pro (see pricing).

Tips

• Make readiness probes check real dependencies, not only that the process is up.
• Set surge and unavailable limits to match your capacity.
• Keep the previous revision available for rollback.

Frequently asked questions

What does a readiness probe do?

It tells Kubernetes whether a pod is ready to receive traffic. Pods that fail are kept out of the Service's endpoints.

Does Kubernetes roll back automatically?

A standard Deployment stops progressing when new pods don't become ready; rolling back is usually triggered by your tooling or an operator.

Can I draw an architecture diagram instead?

Yes. Ask the generator for a cloud architecture diagram of your cluster.

Related