0

I have a question regarding Kubernetes Liveness/Readiness probes configuration.

I have an application developed in netCore 3.1 that at this moment, in production env (version 1.0.0), doesn't have configured health checks. I have implemented the health endpoints in the second release (version 2.0.0) but how can I manage the Kubernetes deployment template file in order to be compliant with version v1 that does not have an endpoint?

If I will deploy my template with probes configured, all container that runs on v1 will fail cause no endpoint are reachable. I would like to understand if I can maintain one deployment yml file that will be compatible with v1 (without health) and v2 (with health).

Here I post an example of my actual deployment yml:

apiVersion: apps/v1
kind: Deployment
metadata:
  namespace: "#{tenant}#-test-app"
  name: "#{tenant}#-test-app"
  labels:
    app: "#{tenant}#-test-app"
    product: "#{tenant}#-test-app"
    app.kubernetes.io/name: "#{tenant}#-test-app"
    app.kubernetes.io/version: "#{server_image_version}#"
    app.kubernetes.io/component: "test-app"
    app.kubernetes.io/part-of: "#{tenant}#-test-app"
    app.kubernetes.io/managed-by: "#{managed_by}#"
spec:
  selector:
    matchLabels:
      app: "#{tenant}#-test-app"
  template:
    metadata:
      labels:
        app: "#{tenant}#-test-app"
    spec:
      containers:
        - name: "#{tenant}#-test-app"
          image: mycontainerregistryurl/test-app:#{server_image_version}#
          ports:
            - containerPort: 80
          envFrom:
            - configMapRef:
                name: "#{tenant}#-test-app-variables-config"
          env:
            - name: DD_AGENT_HOST
              valueFrom:
                fieldRef:
                  fieldPath: status.hostIP
            - name: DD_SERVICE_NAME
              value: "#{tenant}#-test-app"
          securityContext:
            allowPrivilegeEscalation: false
            capabilities:
              drop:
                - NET_RAW
      imagePullSecrets:
        - name: test-registries

server_image_version variable could be used to identify if I have to perform an health check or not.

Thanks in advance, Dave.

DavideP
  • 57
  • 10
  • If you update the image version to v2, the containers will be recreated (with the new image) anyway; so is this really a problem? – Emile Pels May 12 '21 at 06:59

1 Answers1

3

To check the liveness for k8s you can use the command like that, we can define an environement variable and after that on the liveness section we can use the cammand to make an if-else to check the current version and specify what we need to execute on each section .

env:
- name: version
  value: v2
livenessProbe:
  exec:
    command:
    - /bin/bash
    - -exc
    - |
          set +x
          echo "running below scripts"
          if [[ $version -eq "v1" ]]; then
            echo "Running script or command for version 1 "
          else
            echo "Running wget to check the http healthy "
              wget api/v2
          fi

I hope that my idea can help you to resolve your issue .

rassakra
  • 1,062
  • 5
  • 9
  • Hi rassakra, I have posted an example of my deployment, so maybe I could use my "server_image_version" to perform that check you are describing... – DavideP May 12 '21 at 06:54
  • Thanks I test your script a bit readjusted for my specific case and it works ;) the only think I changed is -exc in -exec and wget in curl cause I do not have in my container wget pkg. – DavideP May 12 '21 at 09:16
  • Great , Happy to hear that . – rassakra May 12 '21 at 09:17