Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Creating a fast Kubernetes Development Workflow

Creating a fast Kubernetes Development Workflow

Kubernetes is very powerful in orchestrating where your services are running, but building docker images, writing and updating all the necessary YAML files and applying them to your cluster can still become an annoying maintenance overhead. In this talk you will learn how to create a fast, low friction development workflow with Kubernetes and tools like Telepresence and Forge which enables you to quickly and safely build, test and deploy your services.

Bastian Hofmann

June 25, 2019

More Decks by Bastian Hofmann

Other Decks in Programming


  1. @BastianHofmann Creating a fast Kubernetes Development Workflow Bastian Hofmann

  2. None
  3. Container orchestration platform

  4. Deploy, run and scale your services in isolated containers

  5. No vendor lock in

  6. Runs on

  7. Your laptop

  8. None
  9. Bare metal

  10. Cloud Providers

  11. And if you don't want to install and maintain Kubernetes

  12. Managed Kubernetes

  13. None
  14. Standardized APIs

  15. It works the same everywhere*

  16. It works the same everywhere* *mostly

  17. This talk is about how to use Kubernetes

  18. Not only for production workloads

  19. But in your development workflows

  20. Agenda

  21. Deployment of a micro-service application

  22. Some tools to help with local development of this application

    on Kubernetes
  23. Let's have a look at the sample application

  24. None
  25. OpenStack Cloud LoadBalancer NGINX Ingress Controller NGINX Ingress Controller NGINX

    Ingress Controller web-application web-application MySQL Master MySQL Slave quote-svc quote-svc hello-svc hello-svc
  26. external-dns to create DNS entries automatically

  27. cert-manager to retrieve Let's Encrypt certificates automatically

  28. Database is managed by an Operator

  29. MySQL Operator MySQLCluster MySQL pods MySQL statefulset Kubernetes controller manager

    Discovers Creates Creates Discovers Monitors and manages
  30. LinkerD as a Service Mesh for Telemetry

  31. None
  32. If you are interested in the code and how to

    set it up: https:/ /github.com/syseleven/ golem-workshop
  33. Demo

  34. Writing this YAML files is tedious

  35. YAML files are tied to a specific version and a

    specific environment
  36. Production

  37. Staging

  38. Development

  39. Per Development team

  40. Per branch

  41. Per developer

  42. We need to maintain multiple, very similar YAML files with

    slightly different versions and configuration
  43. "Templating"

  44. Great tools because of standardized Kubernetes API

  45. Helm

  46. None
  47. Allows to install applications

  48. So called "charts"

  49. $ helm install stable/wordpress \ --name my-blog \ --namespace blog

  50. Charts can depend on other charts

  51. Multiple deployments of one chart possible

  52. Different release names

  53. Different namespaces

  54. Configuration with values

  55. None
  56. $ helm install stable/wordpress \ --name my-blog \ --namespace blog

    \ -f my-config-values.yaml
  57. Writing your own charts is fairly easy

  58. Scaffolding to get started

  59. $ helm create quote-svc

  60. $ helm install ./quote-svc \ --namespace dev-bastian \ --name dev-bastian-quote-svc

    \ --values dev.yaml --values bastian.yaml
  61. Demo

  62. Still, for development:

  63. Make a code change

  64. Build docker image

  65. Push docker image

  66. Run helm install/upgrade with new image version

  67. Can this be quicker?

  68. Run everything locally

  69. docker-compose

  70. Duplication of the definition of how to run a container

  71. Inconsistencies

  72. If you have a lot of services, you have to

    run a lot locally
  73. Some services locally, some remote

  74. Service Discovery

  75. Not every service is exposed to the Internet

  76. Shared resources with other developers?

  77. Other options?

  78. Tilt

  79. $ tilt up

  80. Watches for code changes

  81. Rebuilds docker image

  82. Deploys to Kubernetes

  83. Sets up port-forwarding

  84. Can sync changed files directly into a running container

  85. Demo

  86. Another approach

  87. None
  88. Creates a two-way proxy between the Kubernetes cluster and you

  89. $ telepresence T: Starting proxy with method 'vpn-tcp'... @fhgbvx65xg|bash-3.2$ curl

    http://quote-svc/quote | jq '.' [ { "ID": 503, "title": "stefan sagmeister", "content": "<p>...</p>\n", "link": "https://quotesondesign.com/stefan- sagmeister-2/" } ]
  90. Swap a running deployment in the cluster with a local

  91. ... or a locally running docker container

  92. $ telepresence --swap-deployment quote-svc --namespace dev-flow-demo --expose 3000 --run npm

    run debug T: Starting proxy with method 'vpn-tcp',... T: Forwarding remote port 3000 to local port 3000.... > quote-svc@1.0.0 debug /Users/bhofmann/forge_test/quote- svc > nodemon --inspect quote-svc.js [nodemon] watching: *.* [nodemon] starting `node --inspect quote-svc.js` Debugger listening on ws:// d879-4b50-a228-440354cca791 quote svc listening on port 3000!
  93. Demo

  94. Summary

  95. Powerful

  96. Great tooling because of common APIs

  97. Especially great if you have multiple services and don't want

    to run everything locally
  98. I just picked helm, tilt and telepresence. There is more

    for different use-cases.
  99. mail@bastianhofmann.de https:/ /twitter.com/BastianHofmann http:/ /speakerdeck.com/u/bastianhofmann https:/ /github.com/syseleven/golem-workshop