Installation

As an alternative to deploying Kubeflow Pipelines (KFP) as part of the Community distribution, you can also choose to deploy only Kubeflow Pipelines standalone.

You should be familiar with Kubernetes, kubectl, and kustomize.

Note: The placeholder 2.16.1 automatically resolves to the latest Kubeflow Pipelines release (e.g., 2.16.1). Users do not need to manually update this value each time a new release is published.

Deploying Kubeflow Pipelines

1. Deploy the Kubeflow Pipelines development flavor standalone and non-production for first experiments:

export PIPELINE_VERSION=2.16.1

kubectl apply -k "github.com/kubeflow/pipelines/manifests/kustomize/cluster-scoped-resources?ref=$PIPELINE_VERSION"
kubectl wait --for condition=established --timeout=60s crd/applications.app.k8s.io
kubectl apply -k "github.com/kubeflow/pipelines/manifests/kustomize/env/dev?ref=$PIPELINE_VERSION"

2. Deploy Kubeflow Pipelines as part of the community distribution for production ready environments:

This includes proper RBAC and security settings for multi-tenant enterprise environments.

• Install just KFP
• Reproducible GHA workflow
• Install with a single command

3. Deploying Kubeflow Pipelines in Kubernetes Native API Mode for non-production experiments:

Kubeflow Pipelines can be deployed in Kubernetes native API mode, which stores pipeline definitions as Kubernetes Custom Resources instead of using external database storage. This mode provides better integration with Kubernetes native tooling and GitOps workflows.

export PIPELINE_VERSION=2.16.1

kubectl apply -k "github.com/kubeflow/pipelines/manifests/kustomize/cluster-scoped-resources?ref=$PIPELINE_VERSION"
kubectl wait --for condition=established --timeout=60s crd/applications.app.k8s.io
kubectl apply -f https://github.com/cert-manager/cert-manager/releases/download/v1.18.2/cert-manager.yaml
kubectl wait --for=condition=Ready pod -l app.kubernetes.io/instance=cert-manager -n cert-manager --timeout=300s
kubectl apply -k "github.com/kubeflow/pipelines/manifests/kustomize/env/cert-manager/platform-agnostic-k8s-native?ref=$PIPELINE_VERSION"
# Alternatively, for multi-user environments with multiple teams or users requiring isolation and RBAC controls on who can access which pipelines (still not production-ready like the community distribution), you can use the multi-user Kubernetes native mode (requires Istio to be installed, so we strongly recommend using the community distribution instead):
# kubectl apply -k "github.com/kubeflow/pipelines/manifests/kustomize/env/cert-manager/platform-agnostic-multi-user-k8s-native?ref=$PIPELINE_VERSION"

💡 Migration Note: If you are upgrading from a previous version not deployed in Kubernetes native API mode, consider leveraging the migration script to export all existing pipelines and pipeline versions as Kubernetes manifests which can be applied after upgrading Kubeflow Pipelines.

Deploying Kubeflow Pipelines with Pod-to-Pod TLS Enabled

Kubeflow Pipelines can be deployed with pod-to-pod TLS enabled. The API server serves over TLS, and all connecting deployments are mounted with CA certificates. This mode provides enhanced security.

Deploy KFP on a KinD cluster with pod-to-pod TLS enabled using the Makefile target here. The corresponding manifests can be manually accessed here.

Accessing the Kubeflow Pipelines UI

The Kubeflow Pipelines deployment requires approximately 3 minutes to complete.

Standalone / non-production

kubectl port-forward -n kubeflow svc/ml-pipeline-ui 8080:80

Open http://localhost:8080 on your browser to see the Kubeflow Pipelines UI.

Community Distribution

Port-Forward / LoadBalancer / Ingress