Minecraft Operator

A Kubernetes operator for managing PaperMC servers with automatic version management, plugin compatibility solving, and scheduled updates.

Note: This operator is designed for single-instance Minecraft servers. 5-10 minutes of downtime during updates is acceptable by design.

Features

• Automatic Version Management — Four update strategies (latest, auto, pin, build-pin) for both Paper and plugins
• Plugin Compatibility Solver — Constraint solver ensures all plugins work with the selected Paper version
• Scheduled Updates — Cron-based maintenance windows with graceful RCON shutdown
• Declarative Plugin Management — Plugins matched to servers via label selectors
• VolumeSnapshot Backups — Cron-scheduled and on-demand backups with RCON consistency hooks and retention policy
• Self-Managed CRDs — CRDs embedded in the operator binary, applied at startup via server-side apply
• Web UI — Built-in dashboard for monitoring servers and plugins
• Hangar Integration — Automatic plugin downloads from PaperMC Hangar repository
• Direct URL Plugins — Download plugins from any HTTPS URL with optional SHA256 checksum verification

Quick Start

Prerequisites

• Kubernetes 1.27+
• Helm 3.14+

Install the Operator

helm install minecraft-operator oci://ghcr.io/lexfrei/charts/minecraft-operator \
  --create-namespace \
  --namespace minecraft-operator-system

CRDs are embedded in the operator binary and applied automatically at startup. No separate CRD chart or kubectl apply step is needed.

Create a Minecraft Server

First, create an RCON secret:

kubectl create secret generic my-server-rcon \
  --namespace default \
  --from-literal=password="your-rcon-password"

Then apply the server manifest:

apiVersion: mc.k8s.lex.la/v1beta1
kind: PaperMCServer
metadata:
  name: my-server
  labels:
    environment: production
spec:
  updateStrategy: "auto"
  updateSchedule:
    checkCron: "0 3 * * *"
    maintenanceWindow:
      enabled: true
      cron: "0 4 * * 0"
  gracefulShutdown:
    timeout: 300s
  rcon:
    enabled: true
    passwordSecret:
      name: my-server-rcon
      key: password
  podTemplate:
    spec:
      containers:
        - name: minecraft
          resources:
            requests:
              memory: "2Gi"
            limits:
              memory: "4Gi"

Add a Plugin

apiVersion: mc.k8s.lex.la/v1beta1
kind: Plugin
metadata:
  name: essentialsx
spec:
  source:
    type: hangar
    project: "Essentials"
  updateStrategy: "latest"
  updateDelay: 168h
  instanceSelector:
    matchLabels:
      environment: production

The plugin is automatically downloaded and installed on all servers matching the label selector.

See the examples/ directory for more use cases.

Update Strategies

Strategy	Description
latest	Always use newest Paper version (ignores plugin compat)
auto	Solver finds best version compatible with all plugins
pin	Stay on specific version, auto-update to latest build
build-pin	Fully pinned version and build, no automatic updates

See Update Strategies Guide for details.

Architecture

Four controllers work together:

1. Plugin Controller — Fetches plugin metadata from Hangar or direct URLs, runs compatibility solver, updates Plugin status
2. PaperMCServer Controller — Manages StatefulSet and Service, resolves Paper version based on update strategy
3. Update Controller — Executes scheduled updates during maintenance windows with graceful RCON shutdown
4. Backup Controller — Creates VolumeSnapshots with RCON save hooks for data consistency, supports cron scheduling, manual triggers, and retention

For detailed architecture and constraint solver algorithms, see DESIGN.md.

Backups

Enable VolumeSnapshot-based backups:

spec:
  backup:
    enabled: true
    schedule: "0 */6 * * *"     # Every 6 hours
    beforeUpdate: true           # Backup before any update
    retention:
      maxCount: 10               # Keep last 10 snapshots

When RCON is enabled, the operator uses RCON hooks (save-all, save-off, save-on) to flush data to disk and disable auto-save before creating a VolumeSnapshot. Without RCON, snapshots are crash-consistent only.

Trigger a manual backup:

kubectl annotate papermcserver my-server \
  mc.k8s.lex.la/backup-now="$(date +%s)" \
  --namespace minecraft

Requirements: A CSI driver with VolumeSnapshot support must be installed in the cluster.

Helm Configuration

Key values:

crds:
  manage: true          # Apply embedded CRDs at startup (default: true)

webui:
  enabled: true         # Enable built-in Web UI (default: true)
  port: 8082

leaderElection:
  enabled: true         # Enable leader election for HA (default: true)

metrics:
  enabled: true         # Enable Prometheus metrics (default: true)
  port: 8080
  serviceMonitor:
    enabled: false      # Enable ServiceMonitor for Prometheus Operator

See charts/minecraft-operator/values.yaml for all options.

Monitoring

The operator exposes Prometheus metrics at /metrics (port 8080 by default):

• Reconciliation: duration, total count, and error count per controller
• Plugin API: request count, error count, and latency per source
• Solver: invocation count and duration per solver type
• Updates: success/failure count per server update

Enable metrics.serviceMonitor.enabled in Helm values for automatic Prometheus Operator discovery.

Web UI

Access the built-in dashboard:

kubectl port-forward svc/minecraft-operator-webui 8082:8082 \
  --namespace minecraft-operator-system

Open http://localhost:8082/ui

Development

make manifests generate   # Generate CRDs and deepcopy methods
make test                 # Run unit tests with envtest
make lint                 # Run golangci-lint
make run                  # Run operator locally against kubeconfig cluster

This project follows strict TDD methodology. See CLAUDE.md for development standards.

Links

• Documentation Site
• Changelog
• Design Document
• GitHub Issues

Contributing

Contributions are welcome. See GitHub Issues for open tasks.

License

BSD-3-Clause. See LICENSE.