Project is dedicated to making a primitive microservices application. This repository contains all scripts for CI/CD and deploying system.
To create namespace:
kubectl create ns msvc-ns
To set namespace msvc-ns as default:
kubectl config set-context --current --namespace=msvc-ns
To deploy system in minikube:
kubectl apply -f scripts_minikube/
To access endpoint (remove -n flag if you deployed to default namespace):
curl $(minikube service gateway --url -n msvc-ns)
I recommend to create cluster of minimum 2 standard nodes (n1-standard-1). It will give you enough resources to deploy this project with predefined resource requests and limits.
To deploy system in GKE:
kubectl apply -f scripts_gke/
To access endpoint:
curl $(kubectl get svc gateway -o jsonpath='{.status.loadBalancer.ingress[0].ip}')
Built using Jenkins. It should NOT be installed as Docker container (e.g. in GKE environment). The hosting machine must have docker and kubectl installed. If you are going to use docker images from your own docker registry account, then you need to correct hardcoded values (e.g. 'anshelen/microservices-backend') in kubernetes and jenkins files. Also Jenkins is using 'msvc-ns' namespace in a cluster. All files are situated in 'jenkins-kubectl' folder.
- Deploy services in GKE
- Install Jenkins with plugins:
- Call script. It will create service account for Jenkins and print a token to access kubernetes cluster
- Create global credentials with the following ID's:
- 'github-creds' - username/password for git repository
- 'dockerhub-creds' - username/password for docker registry
- 'kubernetes-creds' - secret with 'secret text' type containing generated token
- Create Jenkins global environment variable CLUSTER_URL with url of your
kubernetes master node. You can get it with command:
kubectl cluster-info - Create multibranch pipelines for both backend and gateway services. In "Build Configuration" choose fetching Jenkinsfile using remote file plugin. Set appropriate script path for each file (e.g. jenkins-kubectl/Jenkinsfile-backend) and pick up your git repository. Fields 'Local File' and 'Branch Specifier' should be left blank. Toggle "Periodically if not otherwise run" in "Scan Repository Triggers" to scan main repository for changes every a few minutes.
- Checkout from git
- Build and compile project in isolated docker environment (jdk image)
- Test project
- Compose docker image and push it to docker registry with 'latest' and 'v$BUILD_NUMBER' tags
- Trigger kubernetes cluster to use image with the actual 'v$BUILD_NUMBER' tag
You can set up namespace and resource quotas with command:
kubectl apply -f scripts_env/
To call all following commands in created namespace:
kubectl config set-context --current --namespace=msvc-ns
Add repository with you chart:
helm repo add msvc-repo https://anshelen.github.io/microservices-deploy/
Update local repositories data:
helm repo update
To install chart with 'msvc-project' name:
helm install msvc-project msvc-repo/msvc-chart
Test chart:
helm test msvc-project
Show available options:
helm show values msvc-repo/msvc-chart
To upgrade installation:
helm upgrade msvc-project msvc-repo/msvc-chart --set backend.deployment.name=new-name
- Argument --set can be used multiple times
- To keep previously set options use flag --reuse-values
- You should not modify your cluster using 'kubectl'. All manipulations must be done through Helm
- On default gateway service is of LoadBalancer type (gateway.service.type=LoadBalancer). It is nice for deploy on GKE, but for Minikube it is more suitable to set this option to ClusterIP or NodePort
- Horizontal Pod Autoscalers for backend and gateway deployments are disabled by default. To enable, set backend/gateway.hpa.enabled=true. Keep in mind that you must provide requests.cpu value for deployment. If you prepared environment - it is done for you automatically. To enable or change cpu request manually, set backend/gateway.container.resources.requests.cpu=100m option
- By default a service account is created for your deployments. You can cancel it by specifying serviceAccount.create=false option
Setting pipeline using Helm is mostly the same as described above. All files are located in 'jenkins-helm' folder. Differences:
- Helm 3 should be installed on hosting machine and your chart repository should be added manually (see commands above)
- List commands to obtain token for kubernetes credentials (no more need to
call jenkins-kubectl/jenkins-register.sh script):
helm status msvc-project - Create Jenkins global environment variable HELM_PROJECT with a name of your helm project (in out case it is 'msvc-project')
- If you install helm chart without specifying any options - builds might not
update images due to Helm bug. To
walkaround, upgrade the project manually for the first time:
helm upgrade msvc-project msvc-repo/msvc-chart --set any=null
Files with additional 'package' step are collected in 'jenkins' folder. Docker images are created directly from the tested jar-archives (in above methods they were created from scratch). It minimizes pipeline time. Some mandatory global environment variables for Jenkins were added:
- HELM_CHART - your repo with chart name (e.g. msvc-repo/msvc-chart)
- CLUSTER_NAMESPACE - Kubernetes cluster namespace
This software is licensed under the BSD License. For more information, read the file LICENSE.