Horizontal Pod Autoscaling (HPA)

Overview

This page describes how to configure Horizontal Pod Autoscaling on Appian on Kubernetes for self-managed clients. It allows you to scale replicas of Apache Web Server (httpd) based on observed CPU usage.

Horizontal Pod Autoscaler

The Horizontal Pod Autoscaler (HPA) automatically scales the number of pods in a replication Controller, Deployment, ReplicaSet, or StatefulSet based on observed CPU usage.

Appian on Kubernetes only supports autoscaling for Appian's Apache Web Server (httpd) as it is a stateless component. HPA cannot be configured for Appian's stateful components: Search Server, Zookeeper, Kafka, Data Server, Service Manager, and Webapp.

HPA is configured via the .spec.httpd.hpa fields on Appian custom resources. The following custom resource snippet defines an Appian site with autoscaling enabled for httpd:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
apiVersion: crd.k8s.appian.com/v1beta1
kind: Appian
metadata:
  name: appian
spec:
  httpd:
    # replicas: # Cannot be set when HPA is enabled

    hpa:
      # minReplicas: 1 # Defaults to 1
      maxReplicas: 5

      # The target average CPU utilization (represented as a percentage of
      # requested CPU) over all the pods
      # https://kubernetes.io/docs/tasks/run-application/horizontal-pod-autoscale/#how-does-the-horizontal-pod-autoscaler-work
      targetCPUUtilizationPercentage: 80

Using Different HPA Versions

The current stable version of the HorizontalPodAutoscaler Kubernetes object is in the autoscaling/v1 API. It only includes support for autoscaling based on CPU usage. The beta version, which includes support for scaling on memory and custom metrics, can be found in autoscaling/v2beta2. See the Kubernetes documentation for more information.

The schema of the HorizontalPodAutoscaler API object is different across its API versions. Right now, the Appian operator and Appian custom resource definition (CRD) only support creating HorizontalPodAutoscaler objects using the autoscaling/v1 API version. If you want to take advantage of HPA's newer, beta features, you may manually create HorizontalPodAutoscaler objects using other API versions out-of-band.

Refer to the Kubernetes documentation for examples of how to create HorizontalPodAutoscaler objects using the autoscaling/v2beta1 API version.

The following code snippet details how to point a HorizontalPodAutoscaler object at an Appian site's httpd Deployment using its scaleTargetRef field:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
apiVersion: crd.k8s.appian.com/v1beta1
kind: Appian
metadata:
  name: appian
spec:
  # Enable httpd, but don't set its hpa field
  httpd: {...}
---
apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    # The name of the httpd deployment is the name of the Appian custom resource
    # (appian) plus the name of the stateless component (httpd)
    name: appian-httpd

Feedback