如何使用Kubernetes和Gitlab CI / CD在Google Cloud Platform中部署登台？

Question

I am playing with Docker , Kubernetes , Google Cloud Platform(GCP) and Gitlab recently to achieve CI/CD from commit to staging . So far I have succeeded testing , building and pushing the image to Container registry of Gitlab .

I have a small node and docker application which output 'Hello world' . Also, I have built my docker image in Container registry of Gitlab . At this moment the process is docker-in-docker. I want to push my image from Gitlab container registry to Kubernetes engine in GCP. I have installed both kubectl and gcloud sdk . The Auto DevOps seems to be promising but I want to implement my own .gitlab-ci.yml file.

Here is my .gitlab-ci.yml below:

stages:
  - testing
  - build
  - staging

variables:
  CONTAINER_TEST_IMAGE: registry.gitlab.com/surajneupane55/node-app-
  testing
 CONTAINER_RELEASE_IMAGE: registry.gitlab.com/surajneupane55/node-
 app-testing:latest


test:
  stage: testing
  image: node:boron
  script:
  - npm install
  - npm test


build_image:
  stage: build
  only: [master]
  image: docker:git
  services:
    - docker:dind
  script:
    - docker login -u gitlab-ci-token -p $CI_BUILD_TOKEN 
      registry.gitlab.com/surajneupane55
    - docker build -t $CONTAINER_TEST_IMAGE .
    - docker push $CONTAINER_TEST_IMAGE


staging_site:

//I need help here!!!
//For staging my project in Kubernetes cluster in GCP
//Already created node-app Kubernetes cluster

Please, let me know if my approach is wrong since this is my first learning project with CI/CD.

Answer 1

A simple gitlab-ci.yml file to build and deploy in GKE with Google Container Registry(GCR) . First, we build image and push it to GCR. With valid credentials, we can easily connect the GKE with GCR and deploy.

stages:
  - build
  - deploy

variables:
  CONTAINER_TEST_IMAGE: gcr.io/node-testing-189614/node-testing:latest




build_image:
  stage: build
  only: [master]
  image: google/cloud-sdk
  services:
    - docker:dind

  script:
    - echo "$GOOGLE_KEY_SECRET" > key.json # Google Cloud service accounts
    - gcloud auth activate-service-account --key-file key.json
    - gcloud config set project node-testing-189614
    - gcloud container builds submit -t $CONTAINER_TEST_IMAGE .




deploy_staging:
  image: google/cloud-sdk
  stage: deploy
  script:
    - echo "$GOOGLE_KEY_SECRET" > key.json # Google Cloud service accounts
    - gcloud auth activate-service-account --key-file key.json
    - gcloud config set project node-testing-189614
    - gcloud config set compute/zone europe-west1-b
    - gcloud config set container/use_client_certificate True
    - gcloud container clusters get-credentials node-testing
    - kubectl delete pods --all
    - kubectl apply -f staging.yml

  environment:
    name: staging
    url: http://***.***.**.***:****/ //External IP from Kubernetes
  only:
  - master

Above we delete pods in GKE because we always want to update the image with the latest tag. The best possible solution available so far is to delete the pods and let the staging.yml file creates new one if not available.

Finally staging.yml looks like this:

apiVersion: apps/v1beta1
kind: Deployment
metadata:
  name: node-testing
spec:
  replicas: 2
  template:
    metadata:
      labels:
        app: node-testing
    spec:
      containers:
      - name: node-testing
        image: gcr.io/node-testing-189614/node-testing:latest
        imagePullPolicy: Always
        ports:
        - containerPort: 8080
      imagePullSecrets:
        - name: gcr.io/node-testing-189614/node-testing

Answer 2

Yu do not need the image to actualy be stored in GCR to be able to use it in your GKE, although having it nearby is nice. You need a service account on gcloud so that you have a non expiring auth to GCR (or you need to use gcloud cli to auth to GCR) then just tag the image and push it.

Running it on kubernetes is a different story, and I strongly encurage you to also look at Helm for creating installation charts for your application that can be reused for multiple environments.

Answer 3

You can decouple your configuration from CI to make it more reliable and secure if you follow the GitOps approach.

Please take a look at my answer to another very similar question .

For a more high-level overview, see also:

• Deploy Applications & Manage Releases
• Google Cloud Platform and Kubernetes

---

Disclaimer: I am a Kubernetes contributor and Weaveworks employee. We build open-source and commercial tools that help people to get to production with Kubernetes sooner.