Add Health Probes to your service

By default, Ingress controller will provision an HTTP GET probe for the exposed pods. The probe properties can be customized by adding a Readiness or Liveness Probe to your deployment/pod spec.

With readinessProbe or livenessProbe

apiVersion: extensions/v1beta1
kind: Deployment
metadata:
  name: aspnetapp
spec:
  replicas: 3
  template:
    metadata:
      labels:
        service: site
    spec:
      containers:
      - name: aspnetapp
        image: mcr.microsoft.com/dotnet/core/samples:aspnetapp
        imagePullPolicy: IfNotPresent
        ports:
        - containerPort: 80
        readinessProbe:
          httpGet:
            path: /
            port: 80
          periodSeconds: 3
          timeoutSeconds: 1

Kubernetes API Reference:

Note

  • readinessProbe and livenessProbe are supported when configured with httpGet.
  • Probing on a port other than the one exposed on the pod is currently not supported.
  • HttpHeaders, InitialDelaySeconds, SuccessThreshold aren't supported.

Without readinessProbe or livenessProbe

If the above probes aren't provided, then the Ingress Controller makes an assumption that the service is reachable on the Path specified for backend-path-prefix annotation, or the path specified in the ingress definition for the service.

Default Values for Health Probe

For any property that can't be inferred by the readiness/liveness probe, default values are set.

Application Gateway Probe Property Default Value
Path /
Host localhost
Protocol HTTP
Timeout 30
Interval 30
UnhealthyThreshold 3