LinuxCon China 2018: Microservices, Service Mesh, and CI/CD Pipelines - Making it All Work Together

Microservices come with many advantages for massively scaling applications. CI/CD is a common approach to automating updates to these applications in kubernetes. Strategies such as canary deployments can be simple with a basic website, but this is a difficult task with distributed systems. I would argue that many CI/CD systems are missing some key components such as: advanced routing, observability, and chaos testing.

Service mesh solutions bring these missing components to kubernetes allowing us to shape traffic to specific versions and obtain metrics to better understand the impact of updates.

I will demonstrate how to use common CI/CD tooling such as Brigade, Jenkins, or Spinnaker to drive deployments with Kubernetes. I will show how service mesh technologies such as istio, conduit, or linkerd ease the ability to efficiently deliver and test microservices in Kubernetes.

Transcript

1. Microservices, Service Mesh, and CI/CD Pipelines: Making it All Work Together Brian Redmond, Microsoft
2. Who am I? Brian Redmond • Cloud Architect @ Microsoft (17 years) • Azure Global Black Belt Team • Live in Pittsburgh, PA, USA • Avid marathon runner and outdoors enthusiast • World traveler @chzbrgr71
4. Microservices @chzbrgr71 The answer to all problems.
5. Containers @chzbrgr71 The answer to all problems.
6. Kubernetes @chzbrgr71 The answer to all problems.
11. @chzbrgr71
12. Deployment Strategies • Container Native Development • Continuous Integration • Continuous Deployment • Blue/Green Testing • Canary Testing • A/B Testing @chzbrgr71
13. Blue / Green Testing v1 proxy v2 testing @chzbrgr71
14. Canary Testing v1 proxy v2 80% 20% 100% @chzbrgr71
15. A / B Testing v1 proxy v2 Duplicate for analysis Live Traffic @chzbrgr71
16. Microservices Deployment ui proxy @chzbrgr71
17. What’s missing?
18. advanced routing
19. observability
20. chaos testing
21. Enter Istio • Service discovery and routing • Health checking • Observability • Authentication and authorization • Resiliency • Policy enforcement @chzbrgr71
22. Istio @chzbrgr71
23. What can I do with Istio? advanced routing → route rules, traffic shaping observability → metrics, logs, and tracing chaos testing → fault injection (delays, faults) @chzbrgr71
24. what about a CI/CD tool?
25. @chzbrgr71
26. What is brigade? brigade = event-driven scripting for kubernetes • Encapsulate functions in containers • Run in parallel or serial • Trigger workflow from Github, Docker registry, etc. • Javascript (pipeline as code) • Project config stored as secrets • Well suited for CI/CD pipelines @chzbrgr71
27. @chzbrgr71
28. Demo Time @chzbrgr71
29. web-earth Demo App Architecture api-earthquakes api-fracking api-weather api-flights api-location Azure CosmosDB Azure Kubernetes Service @chzbrgr71
30. Demo - What you will see web-earth api-earthquakes webhook ACR Build helm slack Azure Container Registry Istio api-location version: prod api-location version: canary 100% 50% “virtualservice” 50% @chzbrgr71
31. Summary • Microservices and containers bring added challenges for deployment/testing • Testing in production is a good idea (with canary) • Service Mesh solutions help with: – Observability – Advanced Routing – Chaos Testing • CI/CD can automate the process @chzbrgr71
33. Questions? you can find me at: • Twitter, Github: @chzbrgr71 • brian.redmond@microsoft.com • Demo code: https://github.com/chzbrgr71/os-summit-tokyo • Slides: @chzbrgr71