KubeCon China - Progressive Delivery made easy with Argo Rollouts

Transcript

1. Kevin Dubois Sr Principal Dev Advocate, Red Hat Progressive delivery made easy with Argo Rollouts
2. Kevin Dubois ★ Sr. Principal Developer Advocate at Red Hat ★ Java Champion ★ Technical Lead, CNCF DevEx TAG ★ From Belgium 󰎐 ★ Professional Software Engineer for ~20 years ★ 🗣 Speak English, Dutch, French, Italian youtube.com/@thekevindubois linkedin.com/in/kevindubois github.com/kdubois @kevindubois.com
3. How it started…
6. CI / CD
7. CI / CD Build Test Security Checks Release Deploy Stage Deploy Prod Continuous Integration Continuous Delivery Manual
8. Every 4 months Every week/day/hour 8
9. Continuous Developer Flow Outer loop Inner loop Pull/Merge Request Production Build / Package Code Push Debug Code Review Build Deploy Security Tests Compliance Inner loop Outer loop Developer Test
10. CI - CD - CD Build Test Security Checks Release Deploy Stage Deploy Prod Continuous Integration Continuous Delivery Continuous Deployment Manual Auto
11. GitOps
12. What is GitOps? Treat everything as code Git is the single source of truth Operations through Git workflows
13. CI/CD Engines Jenkins Spinnaker Tekton Concourse CI …... CI/CD versus GitOps Desired State Cluster State Observe State Take Action GitOps Engines ACM, ArgoCD, FluxCD Razee, Faros Desired State Cluster State
14. ArgoCD Sync Monitor Detect drift Take action Argo CD is a declarative, GitOps continuous delivery tool for Kubernetes. Cluster and application configuration versioned in Git Automatically syncs configuration from Git to clusters Drift detection, visualization and correction
15. GitOps Application Delivery Model Push Pull Pull Request Source Git Repository Image Registry Config Git Repository Kubernetes Deploy Monitor Detect drift CD Take action
16. By default: a big bang / all or nothing release
17. Progressive Delivery
18. What is Progressive Delivery? • No Big Bang • Deploy != Release • Metrics • Subset of Users
19. Why Progressive Delivery? • Decreases Downtime • Limits the Tragedy • Deploy & Release to Production faster • Less mocking or setting up unreliable ‘fake’ services
20. Delivery Techniques
21. Blue Green Deployment • All Or Nothing • Quick Rollback
22. Canary Releases • Small Percentage • Increase depending on metrics
23. Dark Launches • Mirroring Traffic • Dark Canaries • Feature Flags
25. The New Pyramid?
26. How to accomplish Progressive Delivery
27. Accomplishing Progressive Delivery with
28. Blue - Green apiVersion: v1 kind: Service metadata: name: my-service labels: app: mystuff spec: ports: - name: http port: 8000 selector: inservice: mypods type: LoadBalancer apiVersion: apps/v1 kind: Deployment metadata: name: mynode-deployment spec: replicas: 1 selector: matchLabels: app: mynode template: metadata: labels: app: mynode spec: containers: - name: mynode image: quay.io/rhdevelopers/mynode:v1 ports: - containerPort: 8000 kubectl label pod -l app=mynode inservice=mypods
29. Canary Releases $ kubectl scale deployment myapp-v1 --replicas=3 $ kubectl scale deployment myapp-v2 --replicas=1
30. Controlling Microservices with a Service Mesh Code Independent (Polyglot) • Chaos: Fault Injection • Resilience: Circuit Breakers • Observability & Telemetry: Metrics and Tracing • Security: Encryption & Authorization • Fleet wide policy enforcement • Intelligent Routing and Load-Balancing • Smarter Canary Releases • Dark Launch
32. Canary Release apiVersion: networking.istio.io/v1alpha3 kind: VirtualService metadata: name: recommendation spec: hosts: - recommendation http: - route: - destination: host: recommendation subset: version-v1 weight: 75 - destination: host: recommendation subset: version-v2 weight: 25
33. Shadowing Traffic apiVersion: networking.istio.io/v1alpha3 kind: VirtualService metadata: name: recommendation spec: hosts: - recommendation http: - route: - destination: host: recommendation subset: version-v1 mirror: host: recommendation subset: version-v2
34. Dark Canary apiVersion: networking.istio.io/v1alpha3 kind: VirtualService metadata: name: recommendation spec: hosts: - recommendation http: - match: - headers: end-user: exact: Alexandra route: - destination: host: recommendation subset: version-v2 - route: - destination: host: recommendation subset: version-v1
35. Automated Progressive Delivery
36. Argo Rollouts
37. Argo Rollouts ArgoCD detects change rollout Monitors Data
38. Rolling out automatically apiVersion: argoproj.io/v1alpha1 kind: Rollout metadata: name: rollouts-demo labels: app: rollouts-demo spec: strategy: canary: steps: - setWeight: 20 - pause: duration: "1m" - setWeight: 50 - pause: duration: "2m" canaryService: rollouts-demo-canary stableService: rollouts-demo-backend trafficRouting: istio: virtualService: name: rollout-vsvc routes: - primary …
39. “Smart” Progressive Delivery Based on Metrics
40. Metrics Based Rollouts strategy: canary: analysis: args: - name: service-name value: rollouts-demo-canary.canary.svc.cluster.local templates: - templateName: success-rate canaryService: rollouts-demo-canary stableService: rollouts-demo-stable trafficRouting: istio: virtualService: name: rollout-vsvc routes: - primary steps: - setWeight: 30 - pause: { duration: 20s } - setWeight: 40 - pause: { duration: 10s } - setWeight: 60 - pause: { duration: 10s } - setWeight: 80 - pause: { duration: 5s } - setWeight: 90 - pause: { duration: 5s } - setWeight: 100 - pause: { duration: 5s }
41. apiVersion: argoproj.io/v1alpha1 kind: AnalysisTemplate metadata: name: success-rate spec: args: - name: service-name metrics: - name: success-rate interval: 10s successCondition: len(result) == 0 || result[0] >= 0.95 failureLimit: 2 provider: prometheus: address: https://internal:demo@prometheus.istio-system.svc.cluster .local:9090 query: | sum(irate(istio_requests_total{ reporter="source", destination_service=~"{{args.service-name}}", response_code!~"5.*"}[30s]) ) Metrics Based Rollouts 41
42. Argo Rollouts Demo @kevindubois 🚀🚀🚀
43. Experiments strategy: canary: steps: - experiment: duration: 1h templates: - name: experiment-baseline specRef: stable weight: 5 - name: experiment-canary specRef: canary weight: 5
44. Final Notes • State is always hard • start with stateless; work with features; non-destructive schema changes; event-driven architectures (use eg. Debezium to work with ‘classic’ DBs). • Step by Step • Embrace GitOps • If you haven’t automatically destroyed something by mistake, you aren’t automating enough • Demos • https://argo-rollouts.readthedocs.io/en/stable/ • https://github.com/kdubois/progressive-delivery • https://github.com/redhat-developer-demos/bubbles-progressive-delivery • https://dn.dev/istio-tutorial
45. Free Developer e-Books & Tutorials! developers.redhat.com/eventtutorials
46. youtube.com/@thekevindubois linkedin.com/in/kevindubois github.com/kdubois @kevindubois.com slides Thank you!