New to Voyager? Please start here.

Scaling Ingress

Replicas

For each Ingress resource, Voyager deploys HAProxy in a Deployment prefixed by voyager- and the name of the Ingress.

This Deployment has .spec.replicas = 1 by default. To start the ingress with the desired number of replicas, use the ingress.appscode.com/replicas annotation.

Note that, Voyager won’t sync with this annotation if there is a HPA controlling the ingress deployment. This regulation is followed so that HPA on ingress deployment does not experience any conflicts.

apiVersion: voyager.appscode.com/v1
kind: Ingress
metadata:
  name: my-app
  namespace: default
  annotations:
    ingress.appscode.com/replicas: '2'
spec:
  defaultBackend:
    service:
      name: my-app
      port:
        number: 80

$ kubectl get deploy voyager-my-app
NAME               DESIRED   CURRENT   UP-TO-DATE   AVAILABLE   AGE
voyager-my-app     2         2         2            2           1d

Horizontal Pod Autoscaling

Kubernetes has the HorizontalPodAutoscaler object for autoscaling pods.

With Horizontal Pod Autoscaling, Kubernetes automatically scales the number of pods in a replication controller, deployment or replica set based on observed CPU utilization (or, with alpha support, on some other, application-provided metrics).

To set up a HorizontalPodAutoscaler for a Voyager HAPRoxy deployment, you can use the kubectl autoscale command or defining a HorizontalPodAutoscaler resource.

kubectl autoscale deployment voyager-my-app --cpu-percent=20 --min=2 --max=10

apiVersion: autoscaling/v1
kind: HorizontalPodAutoscaler
metadata:
  name: voyager-my-app
  namespace: default
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: voyager-my-app
  minReplicas: 2
  maxReplicas: 10
  targetCPUUtilizationPercentage: 20

$ kubectl get hpa
NAME                  REFERENCE                        TARGETS    MINPODS   MAXPODS   REPLICAS   AGE
voyager-my-app        Deployment/voyager-my-app        0% / 20%   2         10        2          1d

Node Autoscaling

If you are using autoscaler to dynamically add and remove nodes, you might be interested in using ingress.appscode.com/node-selector to control which hosts are selected to run HAProxy pods. This is a recommended annotation for HostPort type ingress.