Domain 5 of 6 · Chapter 1 of 2

Deployment Advising

Unlock the complete study guide + 1,040 practice questions across 16 full exams.

Bundled into the existing Professional Cloud Architect premium course — no separate purchase.

Included in this chapter:

  • The three deployment tracks an architect advises
  • The application pipeline: Cloud Build then Cloud Deploy
  • Choosing a deployment strategy
  • Apigee for the API surface and Gemini Cloud Assist
  • Exam-pattern recognition

Deployment strategies: blast radius versus cost and speed

StrategyHow traffic moves to the new versionDowntimeRollback speedCost / complexityNative support
RecreateStop all old, start all newYes, during the swapRedeploy the old versionLowest; one environmentPlain Deployment (Recreate)
Rolling updateReplace instances incrementally, old and new coexistNoRoll the update backwardLow; one environmentGKE / Kubernetes default
Blue-greenCut all traffic over in one switch to a parallel environmentNoFastest; flip back to blueHigh; two full environmentsCloud Run traffic split; manual on GKE
CanarySmall percentage first, verify, then widen progressivelyNoShift the percentage backHighest; needs traffic split and verificationCloud Deploy canary; Cloud Run percentages

Decision tree

What are you deploying?InfrastructureInfrastructure Managermanaged Terraform (IaC)API surfaceApigeeAPI proxy + policiesApplicationBuild the artifact, or promote it?Build + testCloud BuildCI: build steps to imagePromote / releaseCloud DeployCD: pipeline + rollbackPick strategy by riskchoose strategySmallest blast radiusCanary% traffic, verify, widenInstant rollbackBlue-greentwo environments, switchNo 2nd environmentRolling updateincremental; old + new coexistRecreate (accepts downtime) suits batch or windowed internal tools. Gemini Cloud Assist drafts the IaC and aids root-cause throughout.

Cheat sheet

  • Deploy infrastructure, application, and API on three separate tracks
  • Cloud Build builds and tests the artifact, then stops at the registry image
  • Cloud Deploy promotes one release through a defined target sequence
  • Gate production with a Cloud Deploy approval and keep rollback one command away
  • Ship infrastructure with Infrastructure Manager, the managed Terraform service
  • Pick the deployment strategy by risk tolerance against cost and complexity
  • Recreate is the simplest strategy but takes downtime during the swap
  • Rolling update keeps the service up with no second environment
  • Blue-green gives the fastest rollback at the cost of two environments
  • Canary limits blast radius by sending a small traffic slice first
  • Cloud Deploy ships two strategies: standard and canary
  • On Cloud Run, do canary and blue-green by splitting traffic across revisions
  • Front a managed API with an Apigee API proxy, not the raw backend
  • Apigee manages the API lifecycle; a load balancer only routes traffic
  • Use Gemini Cloud Assist as an assistant across the deployment lifecycle
  • Binary Authorization is the deploy-time gate that blocks unsigned images
  • Gate a Cloud Build trigger with --require-approval and the Cloud Build Approver role
  • Fire Cloud Build on an image push with a Pub/Sub trigger on the gcr topic
  • Reuse one cloudbuild.yaml across environments with trigger-specific substitutions
  • Use GKE blue-green node-pool upgrades with a soak duration for quick rollback
  • Protect availability during node upgrades with a PodDisruptionBudget

Unlock with Premium — includes all practice exams and the complete study guide.

Also tested in

References

  1. Infrastructure Manager overview
  2. Cloud Build overview
  3. Cloud Deploy overview
  4. What is Apigee?
  5. Artifact Registry overview
  6. Using Skaffold with Cloud Deploy
  7. Cloud Deploy deployment strategies
  8. Cloud Deploy canary deployment strategy
  9. Cloud Run: rollouts, rollbacks, and traffic migration
  10. Cloud Load Balancing overview
  11. Gemini Cloud Assist overview
  12. Binary Authorization overview