◂ Previous

k8s-spring-boot-helloworld-liveness-readiness-probes

Taking the Kubernetes Spring Boot Hello World Application further. This example adds readiness and liveness probes. Readiness probes helps to know when a Container is ready to start accepting traffic. Liveness probes helps to know when to restart a Container. For example, liveness probes could catch a deadlock, where an application is running, but unable to make progress.

Read more about liveness and readiness probes here.

Prerequisites

• Java IDE (I am using IntelliJ CE)
• Maven
• Docker
• Helm
• Tiller on the Kubernetes cluster

System Configuration at time of test

• macOS Catalina - Version 10.15.3 (19D76)
• IntelliJ CE - Version CE 2019.3
• Maven - Version 3.6.1
• Docker Desktop - Version 2.2.0.3 (42716)
• Kubernetes - v1.15.5
• Helm - v2.14.3

Initial Setup

Adding Spring Boot Actuator to the Application

Add this code to this pom.xml Maven build file adding dependency of Spring Boot Actuator if already not added to your project.

<dependency>
    <groupId>org.springframework.boot</groupId>
    <artifactId>spring-boot-starter-actuator</artifactId>
    <version>2.2.5.RELEASE</version>
</dependency>

Also need to enable endpoints actuator endpoints by adding the following lines to /src/main/resources/application.properties

# Spring Boot Actuator
management.endpoints.enabled-by-default=true

Note: Get the latest release from here

More information about Spring Boot Actuator can be found here.

Adding liveness probe

Add this code to this spring-boot-helloworld-chart/templates/deployment.yaml in the spec.template.spec.containers[0] section

          livenessProbe:
            httpGet:
              path: /actuator/health
              port: {{ .Values.deployment.port }}
            initialDelaySeconds: 30
            periodSeconds: 30
            failureThreshold: 3

This will check to see an HTTP 200 response from the endpoint /actuator/health at the deployment port every 30 seconds (periodSeconds) after an initial delay (initialDelaySeconds) of 30 seconds for a maximum of 3 times (failureThreshold) after which it is going to restart the container for which this liveness probe is added which is the container running the spring boot hello world app.

Adding readiness probe

Add this code to this spring-boot-helloworld-chart/templates/deployment.yaml in the spec.template.spec.containers[0] section

          readinessProbe:
            httpGet:
              path: /actuator/health
              port: {{ .Values.deployment.port }}
            initialDelaySeconds: 15
            periodSeconds: 30
            failureThreshold: 3

This will check to see an HTTP 200 response from the endpoint /actuator/health at the deployment port every 30 seconds (periodSeconds) after an initial delay (initialDelaySeconds) of 15 seconds for a maximum of 3 times (failureThreshold) after which it is going to Pod will be marked container as Unready and no traffic will be sent to it for which this readiness probe is added which is the container running the spring boot hello world app.

Adding ability to deploy same pod again

Add this code to this spring-boot-helloworld-chart/templates/deployment.yaml in the spec.template.metadata section

      annotations:
        timestamp: "{{ date "Mon Jan _2 15:04:05 2006" .Release.Time }}"

The "Mon Jan _2 15:04:05 2006" is the format of the date and the .Release.Time will use the time of the deployment.

To read more about date formats click here. To read more about annotations click here.

Adding Deployment Strategy

Add this code to this spring-boot-helloworld-chart/templates/deployment.yaml in the spec section

  strategy:
    rollingUpdate:
      maxSurge: 1
      maxUnavailable: 0

This tell kubernetes how to replace old Pods by new ones. In this case I am using the RollingUpdate (rollingUpdate). The maxSurge of 1 specifies maximum number of Pods that can be created over the desired number of Pods 1 in this case. The maxUnavailable specifies the maximum number of Pods that can be unavailable during the update process 0 in this case.

Note: Rebuild your docker container using docker build -t spring-boot-helloworld:v1 . as there is a slight change also to the Dockerfile which now calls a procfile on container initialization. We will use this in the future.

Test

Install container using helm command

helm install --name spring-boot-helloworld -f ./spring-boot-helloworld-chart/values.yaml ./spring-boot-helloworld-chart/

Alternatively you can also use upgrade command with the --install flag:

helm upgrade spring-boot-helloworld ./spring-boot-helloworld-chart/ -f ./spring-boot-helloworld-chart/values.yaml --install --namespace default

You can see that the pos is in the initializing state through the terminal and the dashboard:

You can see that the pod is started in about 30 seconds and by going to http://localhost:31000/helloworld/ and following response will show up. The actuator endpoint will also show status as UP:

Now Update App using helm command

helm upgrade spring-boot-helloworld ./spring-boot-helloworld-chart/ -f ./spring-boot-helloworld-chart/values.yaml --install --namespace default

View the new pod getting created through the dashboard and the old pod getting cycled out.

Old pod will be deleted eventually to see only one pod on dashboard:

View the app again in a browser http://localhost:31000/helloworld/:

Cleanup

To stop the container that is running use this command: helm delete --purge spring-boot-helloworld

Useful links

• https://www.baeldung.com/spring-boot-kubernetes-self-healing-apps
• https://www.baeldung.com/kubernetes-helm
• https://medium.com/@pablorsk/kubernetes-helm-node-hello-world-c97d20437abd
• https://docs.spring.io/spring-boot/docs/current/reference/html/production-ready-endpoints.html
• https://kubernetes.io/docs/tasks/configure-pod-container/configure-liveness-readiness-startup-probes/
• https://mvnrepository.com/artifact/org.springframework.boot/spring-boot-starter-actuator
• https://golang.org/src/time/format.go
• https://kubernetes.io/docs/concepts/overview/working-with-objects/annotations/
• https://medium.com/@chunjenchen/helm-upgrade-is-not-recreating-pod-f1813ce8e55a

Next ▸