Ruby conf 2016 - Rails Deploy at StraaS.io

Slide for ruby conf 2016 lightning talk.

Transcript

1. Rails Deploy at StraaS.io @gaga5lala
2. LIVEhouse.in
3. StraaS.io
4. GCP partner
5. Outline 1. Capistrano 2. Docker 3. Kubernetes
6. Sample architecture Computer Instance Computer Instance L O A D B A L A N C E Requests Computer Instance web api payment
7. Capistrano
8. Tools
9. Infrasturcture as code => Infrastructure described as code is flexible, versionable, human-readable, and testable. Chef
10. • Chef server • Chef workstation • Chef node • Role, env, run list • Cookbook ◦ template ◦ recipe ◦ attributes ◦ data bag Chef
11. Chef $ knife bootstrap 192.168.11.22 "sudo chef-client"... $ knife ssh "name:fe-rails-*" "sudo chef-client" -x app -a cloud_v2.public_ipv4
12. A deployment automation tool built on Ruby, Rake, and SSH. $ cap production deploy:check $ cap production deploy branch_name=master $ cap production deploy:bundle $ cap production deploy:compile_assets $ cap production deploy:migrate $ cap production deploy:restart Capistrano
13. Deploy pipeline 1. Create an instance 2. Prepare rails environment (package, configuration, credential) 3. Execute capistrano task 4. Build instance template & Rolling update
14. Docker
15. docker 從入門到放棄 Docker —— 從入門到實踐 Why Docker? - A very light-weight virtaul machine - Efficient - Quickly scale - Build once, deploy everywhere - NO more “it works on my machine”!!
16. Tools
17. Tools
18. Dockerfile
19. Build a image from dockerfile $ cap production docker:image:build:os $ cap production docker:image:build:config $ cap production docker:image:build:web
20. Run a docker container $ cap production docker:start $ cap production docker:stop
21. $ docker exec -it rails_container /bin/bash $ cap production docker:rails:c Debug
22. Push image to docker registry $ docker push gaga5lala/rails:abcde # in Google cloud Platform $ gcloud docker --project my_proj push rails:abcde $ cap production docker:push:web
23. L O A D B A L A N C E web N G I N X web 2 fluentd log Requests rollbar agent web N G I N X web 2 fluentd log rollbar agent Docker containers
24. Deploy pipeline(capistrano) 1. Create an instance 2. Prepare rails environment (package, config, credential) 3. Execute capistrano task 4. Build instance template & Rolling update
25. Deploy pipeline 1. Create an instance 2. Install docker 3. $ cap production docker:build:web 4. $ cap production docker:test:web (request response) 5. $ cap production docker:push:web 6. $ cap production docker:run
26. Scaling? a. Call GCP api? i. snapshot ii. image iii. instance template iv. rolling update b. Use kubernetes!
27. Difference concept • GCP => instance • K8s => container goal: reliable, scalable service
28. Kubernetes
29. Why kubernetes?
30. Docker don’t solve... • A running docker container is not a reliable service • Where to put your docker images? • Resource management & health check • Manage multiple running docker containers • Rolling update & rollback • High availability & load balancing • Auto-scaling • Config management
31. What is Kubernetes Container Cluster Manager, abbreviated as k8s Kubernetes Cluster Deployment Operation Scaling
32. What is Kubernetes • Yaml config files • Specify a desired state • Observe, diff, act - decoupling between application and kernel/os - service should be self-healing - reconciliation between desired state and current state
33. Tools
34. Tools
35. - Pod - Deployment - Service - Ingress - HPA k8s concepts
36. • Running docker container • Yaml config Pod
37. containers: - name: fe-rails-cron image: rails:abcde resources: limits: cpu: 2000m memory: 2000Mi args: - whenever update && cron start livenessProbe: httpGet: port: 80 initialDelaySeconds: 10 Pod
38. containers: - name: fe-rails-cron image: rails:abcde resources: limits: cpu: 2000m memory: 2000Mi args: - whenever update && cron start livenessProbe: httpGet: port: 80 initialDelaySeconds: 10 Pod
39. containers: - name: fe-rails-cron image: rails:abcde resources: limits: cpu: 2000m memory: 2000Mi args: - whenever update && cron start livenessProbe: httpGet: port: 80 initialDelaySeconds: 10 Pod
40. Pod containers: - name: fe-rails-cron image: rails:abcde resources: limits: cpu: 2000m memory: 2000Mi args: - whenever update && cron start livenessProbe: httpGet: port: 80 initialDelaySeconds: 10
41. apiVersion: extensions/v1beta1 kind: Deployment spec: strategy: type: RollingUpdate replicas: 5 template: spec: containers: - name: fe-rails-web image: rails:12345 Deployment
42. apiVersion: extensions/v1beta1 kind: Deployment spec: strategy: type: RollingUpdate replicas: 5 template: spec: containers: - name: fe-rails-web image: rails:12345 Deployment
43. apiVersion: extensions/v1beta1 kind: Deployment spec: strategy: type: RollingUpdate replicas: 5 template: spec: containers: - name: fe-rails-web image: rails:12345 Deployment
44. Bin Packing Rails Rails Nodejs Nodejs Rails Always run with resource (CPU+Mem) limitation Machine Docker Container Loses about 0.4 CPU
45. Service • pod (selector) • service type (external name, clusterIP, nodePort, loadBalancer)
46. Service apiVersion: v1 kind: Service metadata: name: fe-rails spec: selector: app: rails stage: production ports: - port: 80 type: LoadBalancer loadBalancerIP: 1.1.1.1
47. Service apiVersion: v1 kind: Service metadata: name: fe-rails spec: selector: app: rails stage: production ports: - port: 80 type: LoadBalancer loadBalancerIP: 1.1.1.1
48. Service apiVersion: v1 kind: Service metadata: name: fe-rails spec: selector: app: rails stage: production ports: - port: 80 type: LoadBalancer loadBalancerIP: 1.1.1.1
49. Service apiVersion: v1 kind: Service metadata: name: fe-rails spec: selector: app: rails stage: production ports: - port: 80 type: LoadBalancer loadBalancerIP: 1.1.1.1 (X) SSL certificate (X) Domain base routing
50. Service apiVersion: v1 kind: Service metadata: name: fe-rails spec: selector: app: rails stage: production ports: - port: 80 type: NodePort (X) SSL certificate (X) Domain base routing ⇒ Ingress
51. Ingress An Ingress is a collection of rules that allow inbound connections to reach the cluster services. - public IP - tls - backend service - health check
52. apiVersion: extensions/v1beta1 kind: Ingress metadata: name: straas-rails annotations: kubernetes.io/ingress.global-static-ip-name: 1.2.3.4 spec: tls: - secretName: ${STRAAS_CERT} rules: - host: api.straas.io http: paths: - backend: serviceName: rails-api servicePort: 80 - host: web.straas.io Ingress
53. apiVersion: extensions/v1beta1 kind: Ingress metadata: name: straas-rails annotations: kubernetes.io/ingress.global-static-ip-name: 1.2.3.4 spec: tls: - secretName: ${STRAAS_CERT} rules: - host: api.straas.io http: paths: - backend: serviceName: rails-api servicePort: 80 - host: web.straas.io Ingress
54. apiVersion: extensions/v1beta1 kind: Ingress metadata: name: straas-rails annotations: kubernetes.io/ingress.global-static-ip-name: 1.2.3.4 spec: tls: - secretName: ${STRAAS_CERT} rules: - host: api.straas.io http: paths: - backend: serviceName: rails-api servicePort: 80 - host: web.straas.io Ingress
55. Horizontal Pod Autoscaling • Deployment • min/max replica • CPU utilization
56. Horizontal Pod Autoscaling apiVersion: extensions/v1beta1 kind: HorizontalPodAutoscaler metadata: name: fe-rails-api spec: scaleRef: kind: Deployment name: rails-api minReplicas: 2 maxReplicas: 10 cpuUtilization: targetPercentage: 80
57. Horizontal Pod Autoscaling apiVersion: extensions/v1beta1 kind: HorizontalPodAutoscaler metadata: name: fe-rails-api spec: scaleRef: kind: Deployment name: rails-api minReplicas: 2 maxReplicas: 10 cpuUtilization: targetPercentage: 80
58. Horizontal Pod Autoscaling apiVersion: extensions/v1beta1 kind: HorizontalPodAutoscaler metadata: name: fe-rails-api spec: scaleRef: kind: Deployment name: rails-api minReplicas: 2 maxReplicas: 10 cpuUtilization: targetPercentage: 80
59. Commands $ kuberctl create -f rails.yaml $ kuberctl delete -f rails.yaml $ kuberctl replace -f rails.yaml
60. StraaS in K8S ingress.yaml service.yaml deployment.yaml
61. Deploy pipeline(docker) 1. Create an instance 2. Install docker 3. $ cap production docker:build:web 4. Test container running (request response) 5. $ cap production docker:push:web 6. $ cap production docker:run
62. 1. Create k8s cluster 2. $ cap production docker:build:web 3. $ cap production docker:push:web 4. $ cap production k8s:update:yaml 5. $ cap production k8s:db:migrate 6. $ cap production k8s:update Deploy pipeline
63. K8s help you solve !! • A running docker container is not a reliable service • Where to put your docker images? • Resource management & health check • Manage multiple running docker containers • Rolling update & rollback • High availability & load balancing • Auto-scaling • Config management
64. Improvement
65. • Use ruby:2.3-slim as base image • Use k8s secret, get rid of chef • Inject config and credential at docker run • Regression test • Scale matrix: QPS • Use shell script to deploy Improvement
66. Thanks! https://livehouse.in/recruit