Modernizing microservices on Kubernetes with Cloud Run for Anthos/GKE

How can you make your existing microservices apps on Kubernetes to make use of Knative (https://knative.dev) to take advantage of better autoscaling, load balancing, scale-to-zero and many other capabilities on a Google Anthos (https://anthos.dev) cluster.

Presented at Google’s Anthos Day @ KubeCon 2019 San Diego, CA by Ahmet Alp Balkan and Jason (Jay) Smith.

Transcript

1. Modernizing microservices with
   Cloud Run for Anthos
   a.k.a. how to get serverless in your Kubernetes cluster
   Anthos Day – KubeCon 2019, San Diego
   Monday, November 18, 2019
   

   View Slide

2. Ahmet Alp Balkan
   Developer Advocate
   @ahmetb
   Jason (Jay) Smith
   Customer Engineer
   @thejaysmith
   

   View Slide

3. Anthos
   Google Cloud Platform On-Premises Public Cloud
   Anthos GKE
   deployed on VMware
   Hub / Connect
   Anthos Config Management
   Anthos Service Mesh
   GKE Anthos GKE
   deployed on AWS
   Kubernetes
   Cluster
   GCP Marketplace
   Anthos Migrate
   ???
   

   View Slide

4. Anthos
   Google Cloud Platform On-Premises Public Cloud
   Anthos GKE
   deployed on VMware
   Hub / Connect
   Anthos Config Management
   Anthos Service Mesh
   GKE Anthos GKE
   deployed on AWS
   Kubernetes
   Cluster
   GCP Marketplace
   Anthos Migrate
   Cloud Run on Anthos
   

   View Slide

5. Serverless
   

   View Slide

6. Serverless
   Operational
   Model
   Programming
   Model
   No infra management
   No ops for scaling
   Managed security Pay per usage
   (request, etc)
   Service-based Request/event
   driven
   Stateless
   applications
   /
   

   View Slide

7. Serverless > Functions
   Compute
   Data Analytics
   ML & AI
   Database &
   Storage
   Smart
   assistants &
   chat
   DevOps
   Messaging
   

   View Slide

8. Serverless philosophy:
   efficient developers + efficient operators
   

   View Slide

9. Developers care about
   velocity, reproducibility,
   not doing infra
   

   View Slide

10. Cloud Run
    Run any stateless container on
    Google’s fully managed infrastructure
    

    View Slide

11. Container image
    to production URL
    in a few seconds
    Run services in
    any language or
    framework
    Fully-managed,
    rapid autoscaling,
    pay per request
    Cloud Run
    

    View Slide

12. 01
    Demo
    Cloud Run
    

    View Slide

13. ✔ Legacy application deployment
    ✔ Fully managed, rapid autoscaling, scale-to-zero
    ✔ Production-ready, secured (HTTPS) endpoint
    What if you want these on your Kubernetes cluster?
    What we just saw
    

    View Slide

14. Serverless, on your terms
    Cloud Run (fully managed)
    • Serverless dev/operator experience
    • Runs on Google’s infrastructure
    • Pay-per-request
    Cloud Run for Anthos
    • Serverless developer experience
    • Runs in your Anthos/GKE cluster
    next to your Kubernetes workloads
    • Customizable/pluggable for your needs
    

    View Slide

15. GKE
    on GCP
    GKE
    on-prem
    Google
    infra
    Cloud Run for Anthos
    Knative
    GKE (Kubernetes) Knative API
    UI CLI YAML
    Cloud Run
    Developer & Operator
    

    View Slide

16. Knative
    Open source API and implementation that codifies
    "serverless on Kubernetes".
    Adds capabilities to Kubernetes to run stateless
    microservices more effectively.
    Heavily customizable and pluggable.
    Managed for you, with Cloud Run for Anthos.
    Has a strong community, backed by Google,
    Red Hat, IBM, SAP and other contributors.
    knative.dev
    knative.dev
    

    View Slide

17. Kubernetes Memory/CPU based autoscaling (slow)
    Knative
    Rapid, request-oriented autoscaling
    Handles traffic spikes
    Knative enhances Kubernetes
    Autoscaling
    More: https://ahmet.im/blog/knative-better-kubernetes-networking/
    

    View Slide

18. Kubernetes N/A
    Knative
    Scale application to 0, if no requests coming
    Activate (0→1) on the next request
    Knative enhances Kubernetes
    Scale to zero
    More: https://ahmet.im/blog/knative-better-kubernetes-networking/
    

    View Slide

19. Kubernetes Connection-based load balancing
    Knative
    Per-request load balancing
    Traffic splitting (blue/green deployments)
    Knative enhances Kubernetes
    Load Balancing
    More: https://ahmet.im/blog/knative-better-kubernetes-networking/
    

    View Slide

20. 02
    Demo
    Cloud Run
    for Anthos
    

    View Slide

21. What we just saw
    ✔ Same developer/ops experience as the fully-managed Cloud Run
    ✔ Knative installation, managed for you by Cloud Run for Anthos on GKE
    ✔ Traffic splitting, without writing YAML files
    ✔ Knative is still Kubernetes.
    

    View Slide

22. Migrating Kubernetes
    Deployments to Cloud Run
    

    View Slide

23. ✓ Stateless applications (microservices, frontends, event handlers, queue processing)
    ✓ Listens on a port number with HTTP or gRPC.
    ✓ Ideally, doesn't take too long to start up and process requests.
    What’s good with Cloud Run?
    

    View Slide

24. apiVersion: apps/v1
    kind: Deployment
    metadata:
    name: hello-web
    spec:
    replicas: 1
    selector:
    matchLabels:
    app: hello
    tier: web
    template:
    metadata:
    labels:
    app: hello
    tier: web
    spec:
    containers:
    - name: main
    image: gcr.io/google-samples/hello-app:1.0
    resources:
    limits:
    cpu: 100m
    memory: 256Mi
    Kubernetes Deployment Kubernetes Service
    apiVersion: v1
    kind: Service
    metadata:
    name: hello-web
    labels:
    app: hello
    tier: web
    spec:
    type: ClusterIP
    selector:
    app: hello
    tier: web
    ports:
    - port: 80
    targetPort: 8080
    

    View Slide

25. apiVersion: apps/v1
    kind: Deployment
    metadata:
    name: hello-web
    spec:
    replicas: 1
    selector:
    matchLabels:
    app: hello
    tier: web
    template:
    metadata:
    labels:
    app: hello
    tier: web
    spec:
    containers:
    - name: main
    image: gcr.io/google-samples/hello-app:1.0
    resources:
    limits:
    cpu: 100m
    memory: 256Mi
    Kubernetes Deployment Kubernetes Service
    apiVersion: v1
    kind: Service
    metadata:
    name: hello-web
    labels:
    app: hello
    tier: web
    spec:
    type: ClusterIP
    selector:
    app: hello
    tier: web
    ports:
    - port: 80
    targetPort: 8080
    

    View Slide

26. apiVersion: apps/v1
    kind: Deployment
    metadata:
    name: hello-web
    spec:
    replicas: 1
    selector:
    matchLabels:
    app: hello
    tier: web
    template:
    metadata:
    labels:
    app: hello
    tier: web
    spec:
    containers:
    - name: main
    image: gcr.io/google-samples/hello-app:1.0
    resources:
    limits:
    cpu: 100m
    memory: 256Mi
    Kubernetes Deployment Kubernetes Service
    apiVersion: v1
    kind: Service
    metadata:
    name: hello-web
    labels:
    app: hello
    tier: web
    spec:
    type: ClusterIP
    selector:
    app: hello
    tier: web
    ports:
    - port: 80
    targetPort: 8080
    

    View Slide

27. apiVersion: apps/v1
    kind: Deployment
    metadata:
    name: hello-web
    spec:
    replicas: 1
    selector:
    matchLabels:
    app: hello
    tier: web
    template:
    metadata:
    labels:
    app: hello
    tier: web
    spec:
    containers:
    - name: main
    image: gcr.io/google-samples/hello-app:1.0
    resources:
    limits:
    cpu: 100m
    memory: 256Mi
    Kubernetes Deployment Kubernetes Service
    apiVersion: v1
    kind: Service
    metadata:
    name: hello-web
    labels:
    app: hello
    tier: web
    spec:
    type: ClusterIP
    selector:
    app: hello
    tier: web
    ports:
    - port: 80
    targetPort: 8080
    

    View Slide

28. apiVersion: apps/v1
    kind: Deployment
    metadata:
    name: hello-web
    spec:
    replicas: 1
    selector:
    matchLabels:
    app: hello
    tier: web
    template:
    metadata:
    labels:
    app: hello
    tier: web
    spec:
    containers:
    - name: main
    image: gcr.io/google-samples/hello-app:1.0
    resources:
    limits:
    cpu: 100m
    memory: 256Mi
    Kubernetes Deployment Kubernetes Service
    apiVersion: v1
    kind: Service
    metadata:
    name: hello-web
    labels:
    app: hello
    tier: web
    spec:
    type: ClusterIP
    selector:
    app: hello
    tier: web
    ports:
    - port: 80
    targetPort: 8080
    

    View Slide

29. apiVersion: v1
    kind: Service
    metadata:
    name: hello-web
    labels:
    app: hello
    tier: web
    spec:
    type: ClusterIP
    selector:
    app: hello
    tier: web
    ports:
    - port: 80
    targetPort: 8080
    apiVersion: apps/v1
    kind: Deployment
    metadata:
    name: hello-web
    spec:
    replicas: 1
    selector:
    matchLabels:
    app: hello
    tier: web
    template:
    metadata:
    labels:
    app: hello
    tier: web
    spec:
    containers:
    - name: main
    image: gcr.io/google-samples/hello-app:1.0
    resources:
    limits:
    cpu: 100m
    memory: 256Mi
    Kubernetes Deployment Kubernetes Service
    

    View Slide

30. apiVersion: v1
    kind: Service
    metadata:
    name: hello-web
    labels:
    app: hello
    tier: web
    spec:
    type: ClusterIP
    selector:
    app: hello
    tier: web
    ports:
    - port: 80
    targetPort: 8080
    apiVersion: apps/v1
    kind: Deployment
    metadata:
    name: hello-web
    spec:
    replicas: 1
    selector:
    matchLabels:
    app: hello
    tier: web
    template:
    metadata:
    labels:
    app: hello
    tier: web
    spec:
    containers:
    - name: main
    image: gcr.io/google-samples/hello-app:1.0
    resources:
    limits:
    cpu: 100m
    memory: 256Mi
    Kubernetes Deployment Kubernetes Service
    

    View Slide

31. Knative Service = Kubernetes Deployment + Kubernetes Service
    apiVersion: apps/v1
    kind: Deployment
    metadata:
    name: hello-web
    spec:
    replicas: 1
    selector:
    matchLabels:
    app: hello
    tier: web
    template:
    metadata:
    labels:
    app: hello
    tier: web
    spec:
    containers:
    - name: main
    image: gcr.io/[...]
    resources:
    limits:
    cpu: 100m
    memory: 256Mi
    apiVersion: serving.knative.dev/v1alpha1
    kind: Service
    metadata:
    name: hello-web
    spec:
    template:
    spec:
    containers:
    - image: gcr.io/[...]
    resources:
    limits:
    cpu: 100m
    memory: 256Mi
    

    View Slide

32. Knative Service = Kubernetes Deployment + Kubernetes Service
    apiVersion: serving.knative.dev/v1alpha1
    kind: Service
    metadata:
    name: hello-web
    spec:
    template:
    spec:
    containers:
    - image: gcr.io/[...]
    resources:
    limits:
    cpu: 100m
    memory: 256Mi
    apiVersion: apps/v1
    kind: Deployment
    metadata:
    name: hello-web
    spec:
    replicas: 1
    selector:
    matchLabels:
    app: hello
    tier: web
    template:
    metadata:
    labels:
    app: hello
    tier: web
    spec:
    containers:
    - name: main
    image: gcr.io/[...]
    resources:
    limits:
    cpu: 100m
    memory: 256Mi
    

    View Slide

33. Serverless Eventing
    with Cloud Run and Kafka
    

    View Slide

34. I want to create a data driven organization.
    event
    

    View Slide

35. Knative eventing
    Benefits
    ● Declaratively bind between event
    producers and deployed services
    ● Scales from just few events to live
    streams
    ● Custom event pipelines to connect with
    your own existing systems
    

    View Slide

36. Knative Eventing
    

    View Slide

37. Knative Eventing
    

    View Slide

38. View Slide

39. 03
    Demo
    Eventing with Kafka
    on Cloud Run for Anthos
    

    View Slide

40. Wrap-up: Cloud Run on Anthos
    ● Easier deployment and operations for microservices.
    ● Increased productivity and velocity.
    ● Enable your developers and operators.
    

    View Slide

41. Thanks!
    Resources
    ● cloud.run
    ○ anthos.dev
    ● knative.dev
    ○ github.com/ahmetb/cloud-run-faq
    ● gitlab.com/serverlessjay/cloud-run-eventing
    ○ knative.tips
    

    View Slide