Ahmed Baha Eddine Alimi - B23-SD-01 [Lab14 + Bonus Task]

[3llimi]

Lab 14 — Progressive Delivery with Argo Rollouts

What was done:

• Installed and validated Argo Rollouts in dedicated namespace (argo-rollouts), including controller and required CRDs (rollouts.argoproj.io, analysisruns.argoproj.io, analysistemplates.argoproj.io)
• Installed and verified Rollouts CLI plugin (kubectl-argo-rollouts / kubectl argo rollouts) and used it for rollout inspection, promotion, abort, and retry operations
• Refactored Helm workload delivery from standard Deployment flow to Argo Rollout-based progressive delivery in k8s/devops-python:
  • added templates/rollout.yaml
  • added templates/service-preview.yaml for blue-green preview traffic
  • kept strategy selection declarative through Helm values
• Implemented Canary strategy with weighted traffic progression and pause gates for controlled promotion
• Implemented Blue-Green strategy with active/preview services and explicit promotion behavior controls (autoPromotionEnabled)
• Verified rollout behavior end-to-end with runtime commands:
  • rollout status tree (kubectl-argo-rollouts get rollout ...)
  • step progression and pause states
  • manual promote and abort flows
• Resolved real implementation issues encountered during lab execution:
  • fixed NodePort collision by switching lab usage to non-conflicting service setup
  • fixed invalid rendered manifest condition (env: null) by conditional Helm templating
  • fixed rollout pod startup blocker caused by ServiceAccount reference mismatch
  • handled server-side apply field-manager conflicts by cleanup/reinstall strategy before runtime validation
• Documented strategy behavior, trade-offs, and operational command reference in k8s/ROLLOUTS.md

Bonus:

• Implemented automated canary analysis with AnalysisTemplate in templates/analysis-template.yaml
• Wired analysis into canary steps so rollout progression is gated by metric evaluation
• Validated runtime AnalysisRun execution (not only manifest rendering):
  • triggered new rollout revision
  • AnalysisRun was created and executed
  • metric evaluation failed according to policy (failed (2) > failureLimit (1))
  • rollout was automatically aborted and marked Degraded
  • stable ReplicaSet remained healthy and continued serving traffic
• Captured terminal/screenshot evidence for:
  • AnalysisTemplate existence
  • AnalysisRun creation and failed status
  • rollout degraded/aborted state with explicit rollback-protection behavior

Outcome:

• Lab 14 requirements completed for both main implementation (canary + blue-green progressive delivery) and bonus runtime validation (analysis-driven automatic abort/rollback safety).