KodeKloud Notes

Learn how to install KEDA (Kubernetes Event-driven Autoscaling) using its official Helm chart. This guide walks through prerequisites, Helm commands, and a detailed overview of the Kubernetes objects deployed by KEDA for seamless auto-scaling.

Prerequisites

A running Kubernetes cluster (v1.16+)
Helm v3+ installed and configured
Access to the KEDA Helm repository

Note

Ensure you have kubectl access and the correct context set before proceeding.

Add and Install the KEDA Helm Chart

Run the following commands to add the KEDA chart repository, update it, and install KEDA into the keda namespace:

# Add the KEDA Helm repo
helm repo add kedacore https://kedacore.github.io/charts

# Update local Helm repo information
helm repo update

# Install KEDA into namespace "keda"
helm install keda kedacore/keda \
  --namespace keda \
  --create-namespace

Warning

If you already have a keda namespace or a previous installation, use helm upgrade --install to avoid conflicts.

What the Helm Chart Deploys

When you deploy KEDA with Helm, the following Kubernetes resources are created to enable event-driven autoscaling:

Component Type	Resource Name(s)	Purpose
Custom Resource Definitions	`ScaledObject`, `ScaledJob`, `TriggerAuthentication`,<br>`ClusterTriggerAuthentication`, `EventSource`	Extend the API to define KEDA-specific scaling objects.
APIService	`external.metrics.k8s.io`	Exposes external metrics for Horizontal Pod Autoscalers (HPAs).
ServiceAccount	`keda-operator`, `keda-metrics-adapter`, `keda-webhook`	Identifies KEDA processes running in the cluster.
ClusterRoles & RoleBindings	`keda-operator-role`, `keda-metrics-role`,<br>`keda-webhook-role` and bindings	Grants KEDA permissions at cluster and namespace scope.
Deployments	`keda-operator`, `keda-metrics-adapter`, `keda-webhook`	Runs the core controller, metrics server, and admission webhook.

The image outlines KEDA components, including Custom Resource Definitions, API Service, Cluster Roles, Deployments, and Service Accounts, each with a brief description of their functions.

RoleBindings vs. ClusterRoleBindings

KEDA sculpts its permissions with both namespaced and cluster-wide bindings:

RoleBinding
Binds a Role or ClusterRole to the KEDA ServiceAccount within a specific namespace.
ClusterRoleBinding
Binds a ClusterRole to the KEDA ServiceAccount across the entire cluster.

The image is a diagram illustrating the components of KEDA, including Custom Resource Definitions, API Service, Cluster Roles, Deployments, Service Accounts, and Role Bindings, each with a brief description of their functions.

Detailed Breakdown of Deployed CRDs

After installation, you can inspect the following KEDA Custom Resource Definitions:

ScaledObject
Defines how to scale a Deployment or StatefulSet based on external triggers.
ScaledJob
Similar to ScaledObject but for Kubernetes Jobs.
TriggerAuthentication
Stores authentication details for external metric sources.
ClusterTriggerAuthentication
Cluster-scoped version of TriggerAuthentication.
EventSource
Defines custom event sources for autoscaling.

Summary of Installed Objects

CRDs for KEDA’s custom scale definitions
ServiceAccounts for operator, metrics-adapter, and webhook
ClusterRoles & RoleBindings (cluster- and namespace-scoped)
Deployments
- KEDA Operator
- External Metrics Adapter
- Admission Webhook
APIService at external.metrics.k8s.io

With these resources in place, KEDA is ready to drive event-based autoscaling in your Kubernetes cluster. Continue to the hands-on lab to validate each component and start creating ScaledObjects!

Watch Video

Watch video content

Practice Lab

Practice lab