Keldon

Open-source MPP analytics on Kubernetes

Keldon is an open-source Kubernetes operator for Apache Cloudberry. It runs massively parallel analytical workloads with declarative cluster management and self-healing operations. Describe your coordinators, segments, mirrors, storage, and configuration with Kubernetes custom resources; Keldon handles the database lifecycle and keeps the running cluster aligned with your specification.

Key capabilities

  • Day-two automation. Automate bootstrap, failover, recovery, backups, upgrades, and lifecycle operations from one operator-driven control loop.
  • High availability. Automatic coordinator failover through standby promotion and segment mirror recovery.
  • Short-lived SSH certificates. A per-cluster certificate authority signs short-lived SSH certificates, with automatic rotation before expiry and no authorized_keys sprawl.
  • GitOps ready. DatabaseClusterClass and DatabaseConfig let platform teams define standards once and application teams reference them for consistent topology across environments.
  • Backup and restore. Run on-demand full backups to an S3-compatible store with gpbackup, and restore into a cluster with a DatabaseRestore resource.
  • Elastic segments. Increase the segment count to scale out your warehouse. Keldon provisions new pods and redistributes data across the expanded cluster.

Keldon is an open-source project. Star the repository on GitHub to support its development and follow new releases.

Getting started

With Kubernetes 1.29 or later and cert-manager installed, install the Keldon operator with Helm:

helm repo add keldon https://charts.keldon.io
helm install keldon-operator keldon/keldon-operator \
  --namespace keldon-system --create-namespace

Learn more about:

  • Introduction — what Keldon is and how it fits into Kubernetes
  • Getting started — install the operator and deploy your first Cloudberry cluster
  • Custom resources — the API for clusters, images, configuration, backups, and restores
  • Advanced features — high availability, scaling, and database lifecycle operations
  • Operations — running Keldon in production

Under the hood

Apache Cloudberry distributes analytical queries across a coordinator and multiple segments. An optional standby and segment mirrors provide replicas for failover and recovery. Keldon maps these roles to Kubernetes StatefulSets, persistent storage, and Services, and reconciles them against the cluster you declare.

Start with a DatabaseImage describing the database image and a DatabaseCluster referencing it. Use DatabaseClusterClass for reusable topology defaults and DatabaseConfig for database settings and access rules.

The operator manages:

  • cluster initialization and pod orchestration,
  • SSH trust and certificate rotation,
  • coordinator failover and segment recovery,
  • and backups and restores through DatabaseBackup and DatabaseRestore resources.

Read the architecture guide for the details, or the v0.11.1 announcement for the first public release.

Community

Keldon is built in the open. Everyone is welcome, from first-time contributors to production operators. Read the contributing guide or share feedback on GitHub.

The project follows the Code of Conduct. Find news, release notes, and technical guides on the Keldon Blog.

License

Copyright 2026, the Keldon Contributors. Keldon is released under the Apache 2.0 License — see the LICENSE file for details.