* This page is part of the Union.ai v2 docs.
* Full index: https://www.union.ai/docs/v2/union/llms.txt
* Union.ai is a commercial superset of the open-source Flyte project and uses the same `flyte` SDK and CLI, so questions about Flyte are usually answered here too.
* Only refer to the open-source Flyte docs if you are explicitly looking for information about the open-source Flyte project, particularly the open-source Flyte backend. The open-source Flyte v2 docs are at https://www.union.ai/docs/v2/flyte/llms.txt
* These are the v2 docs. v1 is out of date and deprecated. Only refer to v1 docs if asked explicitly about v1. The v1 index is at https://www.union.ai/docs/v1/union/llms.txt

Run the data plane on your own Kubernetes cluster, owning and operating the cluster while Union.ai runs the control plane.

# Self-managed deployment

In a self-managed deployment, you operate the data plane on your own Kubernetes infrastructure.
Union.ai runs the control plane, and you own and operate the Kubernetes cluster the data plane runs on.
On AWS, GCP, and Azure, when you set up from the Union.ai UI, you install an agent that connects out to Union.ai, and Union.ai installs the data plane into your cluster through that connection. Nothing dials in to your cluster.

## AWS, GCP, and Azure: set up from the Union.ai UI

Before you start, [subscribe and sign up](https://www.union.ai/docs/v2/union/deployment/sign-up.md). Then:

1. Provision your cloud resources, using the page for your provider:
   - [AWS](https://www.union.ai/docs/v2/union/deployment/selfmanaged/aws.md)
   - [GCP](https://www.union.ai/docs/v2/union/deployment/selfmanaged/gcp.md)
   - [Azure](https://www.union.ai/docs/v2/union/deployment/selfmanaged/azure.md)
2. [Connect your cluster](https://www.union.ai/docs/v2/union/deployment/selfmanaged/connect-a-cluster.md). You enter your resources' details in the Union.ai UI and install the agent it gives you, and the agent installs the data plane. You skip the Helm install below.

Each provider's page has a **Legacy steps** page under it: the earlier instructions, which install the data plane yourself with Helm. They remain for deployments that were set up that way. For a new deployment on AWS, GCP, or Azure, use the steps above.

## Other providers: set up with Helm

The UI flow covers AWS, GCP, and Azure today. On OCI, CoreWeave, Crusoe, and Nebius, and on generic Kubernetes (on-premises or any other environment), you provision the infrastructure and install the data plane with Helm yourself. Existing AWS, GCP, and Azure deployments that were set up with Helm follow this path too:

1. Review the [platform architecture](https://www.union.ai/docs/v2/union/deployment/platform-architecture.md) to understand the control plane, data plane operators, and security model.
2. Review the [infrastructure recommendations](https://www.union.ai/docs/v2/union/deployment/selfmanaged/infrastructure-recommendations.md) for the sizing model, networking and IP planning, identity, and the [scaling constraints](https://www.union.ai/docs/v2/union/deployment/selfmanaged/infrastructure-recommendations/scaling-constraints.md) to plan around.
3. Provision your infrastructure using the page for your provider:
   - [OCI](https://www.union.ai/docs/v2/union/deployment/selfmanaged/infrastructure-recommendations/oci.md)
   - [Generic Kubernetes](https://www.union.ai/docs/v2/union/deployment/selfmanaged/infrastructure-recommendations/generic.md) (on-premise or any S3-compatible environment)
   - [CoreWeave](https://www.union.ai/docs/v2/union/deployment/selfmanaged/infrastructure-recommendations/coreweave.md)
   - [Crusoe](https://www.union.ai/docs/v2/union/deployment/selfmanaged/infrastructure-recommendations/crusoe.md)
   - [Nebius](https://www.union.ai/docs/v2/union/deployment/selfmanaged/infrastructure-recommendations/nebius.md)
4. [Deploy the data plane](https://www.union.ai/docs/v2/union/deployment/selfmanaged/deploy.md) once your infrastructure is ready. CoreWeave, Crusoe, and Nebius have their deploy steps on their own pages.
5. [Upgrade the data plane](https://www.union.ai/docs/v2/union/deployment/selfmanaged/upgrade.md) when a new chart version is released. You own the chart version, its CRDs, and your values file.

## Configuration

After initial setup, configure platform features on your cluster. Where a page sets values on the `dataplane` Helm chart, it is for a data plane installed with Helm: on one connected from the Union.ai UI, the agent manages the chart and its values (see [Configuration](https://www.union.ai/docs/v2/union/deployment/selfmanaged/configuration.md)).

- [Image builder](https://www.union.ai/docs/v2/union/deployment/selfmanaged/configuration/image-builder.md)
- [Multi-cluster](https://www.union.ai/docs/v2/union/deployment/selfmanaged/configuration/multi-cluster.md)
- [Node pools](https://www.union.ai/docs/v2/union/deployment/selfmanaged/configuration/node-pools.md)
- [Monitoring](https://www.union.ai/docs/v2/union/deployment/selfmanaged/configuration/monitoring.md)
- [Persistent logs](https://www.union.ai/docs/v2/union/deployment/selfmanaged/configuration/persistent-logs.md)
- [Data retention](https://www.union.ai/docs/v2/union/deployment/selfmanaged/configuration/data-retention.md)
- [Namespace mapping](https://www.union.ai/docs/v2/union/deployment/selfmanaged/configuration/namespace-mapping.md)
- [Secrets management](https://www.union.ai/docs/v2/union/deployment/selfmanaged/configuration/union-secrets.md)

## Reference

- [Helm chart reference](https://www.union.ai/docs/v2/union/deployment/selfmanaged/helm-chart-reference.md) for the chart values available when you install with Helm
- [Kubernetes access controls](https://www.union.ai/docs/v2/union/deployment/selfmanaged/kubernetes-rbac.md) for RBAC configuration details

{{< subpage-cards >}}

## Subpages

- [Provision your AWS resources](https://www.union.ai/docs/v2/union/deployment/selfmanaged/aws.md) - Create the EKS cluster, S3 bucket, ECR repository and IAM roles that an AWS cluster pool needs.
  - Agent-assisted provisioning
  - Auto-detect
  - agents
  - claude
  - codex
  - hermes
  - opencode
  - pi
  - Prerequisites
  - 1. Set names and verify your AWS identity
  - 2. Create the EKS cluster
  - 3. Create the S3 bucket
  - 4. Create the private ECR repository
  - 5. Create the system and task IRSA roles
  - 6. Grant S3 and Secrets Manager access
  - 7. Grant repository access
  - 8. Collect the values for connecting your cluster
  - Important behavior and cleanup
  - Next steps
- [Provision your GCP resources](https://www.union.ai/docs/v2/union/deployment/selfmanaged/gcp.md) - Create the GKE cluster, Cloud Storage bucket, Artifact Registry repository, and service accounts that a GCP cluster pool needs.
  - Prerequisites
  - 1. Set your project and names
  - 2. Enable the APIs
  - 3. Create the private network
  - 4. Create the bucket and image repository
  - 5. Create the service accounts and grant access
  - 6. Create the GKE cluster
  - 7. Bind the cluster to the service accounts
  - 8. Collect the values for connecting your cluster
  - Important behavior and cleanup
  - Next steps
- [Provision your Azure resources](https://www.union.ai/docs/v2/union/deployment/selfmanaged/azure.md) - Create the AKS cluster, storage account, container registry, Key Vault, and managed identities that an Azure cluster pool needs.
  - Prerequisites
  - 1. Sign in and choose names
  - 2. Register the resource providers
  - 3. Create the resource group and network
  - 4. Create the storage account and container
  - 5. Create the container registry and Key Vault
  - 6. Create the managed identities and Log Analytics workspace
  - 7. Create the AKS cluster
  - 8. Federate the identities to the Kubernetes service accounts
  - 9. Grant access
  - 10. Check the resources and collect the values
  - Important behavior and cleanup
  - Next steps
- [Connect your cluster](https://www.union.ai/docs/v2/union/deployment/selfmanaged/connect-a-cluster.md) - Connect your cluster to Union.ai, so your workloads run on your own infrastructure instead of your machine.
  - What you'll need
  - What this installs
  - 1. Create the cluster pool
  - AWS
  - GCP
  - Azure
  - 2. Register the cluster
  - AWS
  - GCP
  - Azure
  - 3. Install the agent
  - 4. Wait for the data plane
  - 5. Run a workload on the cluster
  - Next steps
- [What Union.ai installs in your cluster](https://www.union.ai/docs/v2/union/deployment/selfmanaged/whats-installed.md) - The cloud resources and in-cluster services a self-managed data plane runs on, the permissions they hold, and what they cost while idle.
  - Cloud resources you create
  - Services Union.ai installs
  - Permissions
  - What it costs when nothing is running
  - Lower your baseline
  - See it yourself
- [Infrastructure recommendations](https://www.union.ai/docs/v2/union/deployment/selfmanaged/infrastructure-recommendations.md) - Size the data plane, check prerequisites, and plan control plane connectivity before you install.
  - Sizing model
  - Workload signals → size mapping
  - Pre-installation checklist
  - Kubernetes cluster
  - Kubernetes versions
  - Node pools
  - Networking
  - VPC and subnet sizing
  - Public vs. private subnets
  - NAT gateway requirements
  - Identity and workload binding
  - Common identity pitfalls
  - Object storage
  - Container image registry
  - Logging
  - Secrets
  - Control-plane connectivity
  - Cloud provider setup
  - Next
- [Deploy the data plane](https://www.union.ai/docs/v2/union/deployment/selfmanaged/deploy.md) - Install the data plane Helm chart on AWS, GCP, Azure, OCI, or generic Kubernetes.
  - Assumptions
  - Prerequisites
  - Deploy the Union.ai operator
  - Next: manage your cluster and pools
  - Upgrades and maintenance
  - GPU cloud deployment guides
- [Upgrade the data plane](https://www.union.ai/docs/v2/union/deployment/selfmanaged/upgrade.md) - Keep a self-managed data plane current, including the chart, its CRDs, and your values file.
  - What you maintain
  - The supported setup
  - Before you upgrade
  - Upgrade
  - Roll back
  - Automating upgrades
  - One-time migrations
  - Upgrade to 2026.10.1 or later
- [Configuration](https://www.union.ai/docs/v2/union/deployment/selfmanaged/configuration.md) - Configure Union features on your own cluster.
- [Helm chart reference](https://www.union.ai/docs/v2/union/deployment/selfmanaged/helm-chart-reference.md) - The full set of Helm values available across the Union charts.
- [Kubernetes access controls](https://www.union.ai/docs/v2/union/deployment/selfmanaged/kubernetes-rbac.md) - The Kubernetes RBAC the Helm chart applies, and what each role can reach.
  - Service accounts
  - Low-privilege mode vs. standard mode
  - Running in standard mode
  - Multi-namespace mode
  - Roles in every mode
  - `union-system-secret`
  - `operator-system`
  - Low-privilege mode
  - Standard mode
  - ClusterRoles in every mode
  - Service access
  - `operator/operator-proxy`
  - Leaseworker
  - Pod webhook

---
**Source**: https://github.com/unionai/unionai-docs/blob/main/content/deployment/selfmanaged/_index.md
**HTML**: https://www.union.ai/docs/v2/union/deployment/selfmanaged/
