Welcome

Welcome to the Orkestra documentation — the complete reference for the world’s first zero‑code Kubernetes operator runtime.

Orkestra turns CRDs into operators. This documentation explains how.

What Is Orkestra?

Orkestra is a declarative, zero‑code Kubernetes operator runtime. You declare what you want in a YAML file — a Katalog — and Orkestra manages the full lifecycle of your CRDs: create, reconcile, drift‑correct, delete.

No Go. No code generation. No controller boilerplate.

Why Orkestra?

CRDs were meant to make Kubernetes extensible. But to turn a CRD into an operator, you still write Go code — controllers, informers, reconcilers, finalizers, events, metrics. Every operator repeats the same patterns. Every team rebuilds the same infrastructure.

Orkestra removes that barrier. You write a Katalog. Orkestra builds the operator.

Your CRD is enough.

How It Works

Orkestra follows the GitOps pattern of using YAML as the source of truth for operator behavior.

1. You declare a Katalog — a YAML file that describes your CRDs and the resources they should create (Deployments, Services, Secrets, ConfigMaps, Jobs, CronJobs).

2. You apply your CRD — Orkestra detects the CRD, creates informers, starts workers, and begins watching for Custom Resources.

3. You create a Custom Resource — Orkestra reconciles it, creates the declared resources, adds finalizers, and enables drift correction.

4. You update or delete the CR — Orkestra reconciles changes, corrects drift, or cleans up child resources.

For a quick overview, see the Getting Started guide.

Architecture

Orkestra is implemented as a Kubernetes runtime that watches your CRDs and reconciles according to your Katalog. It runs as a single binary — locally or in the cluster — and manages all your CRDs in one process.

• One runtime, any number of CRDs — each CRD gets its own informer, worker pool, and queue
• Dependency‑aware startup — declare dependsOn, Orkestra starts CRDs in topological order
• Built‑in observability — health endpoints, Prometheus metrics, ork status CLI
• Per‑CRD tuning — configure workers, resync intervals, and queue depth per CRD
• Drift correction — resources with reconcile: true are automatically corrected on every reconcile
• No programming language required — just YAML

For additional details, see the Architecture Overview.

Quick Start

# Install Orkestra
brew install orkspace/tap/ork
# or
curl -sSL https://raw.githubusercontent.com/orkestra-sh/orkestra/main/install.sh | bash

For a step‑by‑step walkthrough, see the Getting Started Guide.

Features

• Zero‑code operators — No Go, no Python, no controller boilerplate. Just YAML.
• Any CRD, any resource — Works with custom CRDs and built‑in Kubernetes resources (Pods, Deployments, Secrets)
• Declarative templates — Use Go templates to reference CR spec fields ({{ .spec.image }})
• Conditional provisioning — Create resources only when conditions are met (when: )
• Dependency ordering — Declare dependsOn, Orkestra starts CRDs in the right order
• Drift correction — Resources with reconcile: true are automatically corrected on every reconcile
• Admission policy — Declare validation and mutation rules in the Katalog. No webhook server to run, no TLS to manage, no Go to write.
• Multi‑version CRDs — Declarative conversion paths between CRD versions. No conversion functions.
• Per‑CRD workers — Each CRD gets its own worker pool, queue, and resync interval
• Built‑in observability — Health and info endpoints (/katalog/{crd}/health and /katalog/{crd}), Prometheus metrics(/metrics), ork status
• Komposer composition — Compose Katalogs from files, Helm charts, and remote URLs
• Multi‑source authentication — Support for bearer tokens, GitHub tokens, and basic auth
• Leader election — High availability with warm caches on all replicas
• Graceful shutdown — Drains workers, stops informers, releases leader lease

Development Status

Orkestra is being actively developed. The core runtime is stable and ready for testing. New features are being added regularly.

• Current version: v0.1.0 (alpha)
• Roadmap: See Roadmap
• Releases: Available on GitHub

Adoption

Orkestra is designed for:

• Platform engineers — build internal developer platforms without writing operators
• SREs — manage infrastructure with declarative YAML, not Go code
• Infrastructure teams — replace dozens of operators with one runtime
• Application developers — define their own CRDs without learning Kubernetes controller patterns

Early adopters are using Orkestra to manage:

• Namespace provisioners with quotas and network policies
• Database operators (PostgreSQL, MongoDB) with secrets and backups
• Application operators (Deployment + Service + ConfigMap) with conditional public exposure
• Built‑in resource governance (Pod, Deployment, Secret) with health monitoring

Documentation

Community

• GitHub Issues — report bugs, request features
• Discussions — ask questions, share ideas

Built with ❤️ for the Kubernetes ecosystem. 🎼