Skip to main content

Kubernetes infrastructure

Last updated on

An infrastructure definition links a Harness environment to the physical Kubernetes cluster where you deploy your service. You create both at Deployments > Environments and reuse them across pipelines and stages.


Create an environment

  1. Go to Deployments, expand the drop-down, and select Environments.
  2. Select + New Environment.
  3. Enter a name and select an environment type: Pre-Production or Production. The type affects governance policies and deployment dashboards.
  4. Select Save.

Harness opens the environment detail page with three tabs: Configuration, Infrastructure Definitions, and References.


Create an infrastructure definition

Infrastructure definitions live inside an environment. Each definition links the environment to a specific cluster and namespace.

  1. Open the environment and select the Infrastructure Definitions tab.
  2. Select + Create Infrastructure.
  3. In the first dialog, select Kubernetes.
  4. Select the infrastructure type that matches your cluster provider; the sections below describe each type.
  5. Enter a name; Harness auto-suggests one, and you can change both the name and ID.
  6. Under Storage, choose Inline or a Git repository.
  7. Fill in the Cluster Details under Provisioner Configuration.
  8. Configure Simultaneous deployments and Scope to Specific Services as needed.
  9. Select Save.

Kubernetes

Use this type to connect directly to any Kubernetes cluster, including OpenShift clusters.

FieldRequiredDescription
ConnectorYesA Harness Kubernetes cluster connector.
NamespaceYesThe namespace to deploy into, for example default. Supports expressions and runtime input.
Release NameNoDefaults to release-<+INFRA_KEY_SHORT_ID>. Harness uses this to track releases on the cluster via a ConfigMap.
Delegate-based credentials

If your delegate runs inside the target cluster, configure the connector to use Inherit from Delegate. The delegate already has the service account permissions it needs.

Google Kubernetes Engine

Use this type to connect to GKE clusters via a GCP connector.

FieldRequiredDescription
ConnectorYesA Harness GCP connector.
ProjectNoThe GCP project that contains the cluster. When set, the Cluster dropdown lists only clusters in that project.
ClusterYesThe name of the target GKE cluster. Supports runtime input.
NamespaceYesThe namespace to deploy into. Supports expressions and runtime input.
Release NameNoDefaults to release-<+INFRA_KEY_SHORT_ID>.
GCP Workload Identity

If the delegate runs on a GKE cluster with Workload Identity enabled and the GCP connector is set to inherit credentials from the delegate, Harness uses Workload Identity automatically.

Microsoft Azure

Use this type to connect to AKS clusters via an Azure connector.

FieldRequiredDescription
ConnectorYesA Harness Azure connector.
Subscription IDYesThe Azure subscription that contains the AKS cluster.
Resource GroupYesThe resource group that contains the AKS cluster.
ClusterYesThe name of the target AKS cluster. Supports runtime input.
NamespaceYesThe namespace to deploy into.
Release NameNoDefaults to release-<+INFRA_KEY_SHORT_ID>.

Elastic Kubernetes Service

Use this type to connect to EKS clusters via an AWS connector. Harness uses aws-iam-authenticator to generate the kubeconfig.

FieldRequiredDescription
ConnectorYesA Harness AWS connector.
ClusterYesThe EKS cluster in the format <region>/<cluster-name>, for example ap-south-1/my-cluster. Supports runtime input.
NamespaceYesThe namespace to deploy into.
Release NameNoDefaults to release-<+INFRA_KEY_SHORT_ID>.

Rancher

Use this type to connect to Rancher-managed clusters. You need the Rancher endpoint URL and a bearer token. The Rancher account associated with the token must have the Cluster Owner role or a global permission that enables cluster administration.

FieldRequiredDescription
ConnectorYesA Harness Rancher connector.
ClusterYesThe name of the target Rancher-managed cluster. Supports runtime input.
NamespaceYesThe namespace to deploy into. Supports runtime input.
Release NameNoDefaults to release-<+INFRA_KEY_SHORT_ID>.

Namespace and release name

You can reference the infrastructure definition's namespace in your Kubernetes manifests:

  • CEL: ${{infra.namespace}}
  • JEXL: <+infra.namespace>

This lets you write manifests that work across environments without hardcoding a namespace. If you omit the namespace field from a resource in your manifest, Harness uses the namespace from the infrastructure definition automatically.

The Release Name defaults to release-<+INFRA_KEY_SHORT_ID>. Harness creates a ConfigMap with this name to track all resources belonging to a release. The name must be unique across the cluster and must start with an alphabetic character (Kubernetes RFC-1035 requirement, hence the release- prefix).


Simultaneous deployments

By default, Harness queues deployments to the same infrastructure definition to prevent conflicts; the UI shows Simultaneous deployments: False. Set this to True on the infrastructure definition to allow concurrent deployments to the same target.


Scope to specific services

By default, an infrastructure definition is available to any service in the environment. Enable Scope to Specific Services to limit the infrastructure definition so it only appears when deploying a specific set of services. This prevents teams from accidentally targeting the wrong infrastructure.


Next steps