Cloud Native with ODP — Kubernetes Overview
This feature will be included in ODP 1.3.2.0 as a Tech Preview, currently in qualification. It is available for early enterprise testing.
Interested in early access? Contact our team to join the enterprise early access program.
The Vision: One Control Plane for Hadoop and Kubernetes
ODP is built on the principle that your operations team should not have to manage two entirely separate worlds — a traditional Hadoop cluster on one side, and a Kubernetes platform on the other — using different tools, different access controls, and different monitoring systems.
With Ambari 2.8.2.0, ODP introduces the Kubernetes Manager View: a plugin that extends the Ambari control plane to orchestrate Kubernetes workloads alongside your existing HDFS, YARN, Hive, HBase, and other cluster services. The same Ambari UI your team already uses for cluster administration gains the ability to deploy, configure, upgrade, and monitor Kubernetes-hosted applications — today, Trino and Apache Superset — using Helm charts.
This is not a reimplementation of Kubernetes. Ambari does not replace kubectl or your existing Kubernetes tooling. What it does is provide a governed, integrated deployment path that wires Kubernetes workloads into ODP's security model (Kerberos, Ranger, LDAP) automatically, eliminating the manual integration work that typically makes deploying analytics tools on Kubernetes a multi-week project.
What Problem Does This Solve?
Modern data platforms increasingly need tools that are better suited to container-native deployment than to traditional cluster services:
- Trino (formerly PrestoSQL) benefits from elastic horizontal scaling of workers, which Kubernetes handles natively.
- Apache Superset is a web application that is straightforward to run as a containerized service.
Without ODP's Kubernetes integration, connecting these tools to a Kerberos-secured Hadoop cluster requires manual keytab management, Ranger policy configuration, LDAP wiring, and ongoing synchronization of configuration between systems. A change in the Hive Metastore URI, for example, must be reflected in both the cluster config and the Helm values for Trino.
The Ambari Kubernetes Manager View materializes ODP cluster configuration and security settings directly into Helm chart values at deployment time. Your Kubernetes workloads stay synchronized with the cluster they depend on.
Architecture Overview
┌─────────────────────────────────────────────────────────┐
│ Ambari Server (2.8.2.0) │
│ │
│ ┌─────────────────┐ ┌──────────────────────────┐ │
│ │ Cluster Mgmt │ │ Kubernetes Manager View │ │
│ │ (HDFS, YARN, │ │ (Tech Preview) │ │
│ │ Hive, HBase, │ │ - Helm chart deployment │ │
│ │ Ranger, …) │ │ - GitOps / Flux status │ │
│ └────────┬────────┘ └────────────┬─────────────┘ │
│ │ │ │
└───────────┼────────────────── ────────┼────────────────── ┘
│ │
┌──────▼──────┐ ┌────────▼────────┐
│ ODP Cluster │ │ Kubernetes / │
│ (HDFS, YARN,│◄────────► OpenShift │
│ Hive Meta, │ shared │ │
│ Kerberos, │ infra │ ┌───────────┐ │
│ Ranger, │ │ │ Trino │ │
│ LDAP) │ │ ├───────────┤ │
└──────────────┘ │ │ Superset │ │
└─────────────────┘
Ambari manages both sides. Security configuration (Kerberos realm, keytabs, Ranger policies, LDAP/AD connection) is defined once and applied consistently across both traditional cluster services and Kubernetes-deployed workloads.
What Is Deployed on Kubernetes
In ODP 1.3.2.0 (Tech Preview), Ambari can deploy and manage the following Kubernetes workloads:
| Workload | Version | Description |
|---|---|---|
| Trino | Current | Distributed SQL query engine with Iceberg/Hive Metastore connectivity |
| Apache Superset | 4.1.4 | Business intelligence and data visualization platform |
Both are deployed via Helm charts. Ambari handles chart lifecycle management (install, upgrade, rollback, uninstall) and integrates each workload with the cluster's security infrastructure.
Prerequisites
Before using the Kubernetes Manager View, ensure the following prerequisites are met:
Kubernetes Cluster
- Kubernetes 1.20 or later, or OpenShift 4.8 or later
- Helm 3.x installed and accessible from the Ambari server
- A dedicated namespace for ODP-managed workloads (recommended:
odp-apps) - A service account with sufficient permissions to create deployments, services, configmaps, and secrets in that namespace
- Persistent volume support (for Superset's metadata database)
- Ingress controller or LoadBalancer support for external access
ODP Cluster
- Ambari 2.8.2.0 or later (the Kubernetes Manager View is bundled as a plugin)
- Kerberos enabled (the integration is designed for secured clusters)
- Ranger enabled and configured
- Hive Metastore running (required for Trino connectivity)
- Network connectivity between Kubernetes worker nodes and ODP cluster nodes (Hive Metastore port, HDFS NameNode, KDC)
Network
- Kubernetes nodes must be able to reach the Kerberos KDC
- Kubernetes nodes must be able to reach the Hive Metastore (default: port 9083)
- Kubernetes nodes must be able to reach HDFS NameNode and DataNodes if Trino reads data directly from HDFS
Tech Preview Guidance
The Kubernetes integration is fully functional and suitable for evaluation and pre-production use. Specifically:
What is stable:
- Helm chart deployment and lifecycle management from Ambari
- Kerberos keytab provisioning for Trino
- Ranger authorization for Trino queries
- Hive Metastore / Iceberg catalog connectivity from Trino
- Superset connectivity to Trino and Hive
What may evolve:
- The Ambari Kubernetes View UI layout and configuration forms
- Helm values schema for Trino and Superset charts
- GitOps/Flux integration configuration syntax
Known limitations:
- YARN resource quotas are not enforced for Kubernetes workloads (Trino has its own resource management)
- Atlas lineage is not captured for queries routed through Trino in this release
- High availability for Superset is not yet configured through Ambari (manual Helm override possible)
For production deployments, evaluate the integration thoroughly in a staging environment and subscribe to ODP release notes for updates to this feature.