Helm Chart for PDB Operator

Getting Started   •   Configuration   •   Contributing

---

Helm chart for PDB Operator, a Kubernetes operator that automatically creates and manages PodDisruptionBudgets based on PDBPolicy custom resources. For full documentation, visit pdboperator.io.

Prerequisites

• Kubernetes >= 1.26
• Helm >= 3.x
• cert-manager (required when webhooks are enabled)

Quick Start

# Install from OCI registry
helm install pdb-operator oci://ghcr.io/pdb-operator/charts/pdb-operator \
  --namespace pdb-operator-system \
  --create-namespace

# Verify the operator is running
kubectl get pods -n pdb-operator-system

Upgrading from 0.4.x or earlier? 0.5.0 changes how the PDBPolicy CRD is managed and needs a one-time step. See UPGRADING.md.

Create a PDBPolicy:

apiVersion: availability.pdboperator.io/v1alpha1
kind: PDBPolicy
metadata:
  name: my-pdb-policy
spec:
  availabilityClass: standard
  workloadSelector:
    matchLabels:
      app: my-app

availabilityClass is one of non-critical, standard, high-availability, mission-critical, or custom. Both availabilityClass and workloadSelector are required.

Configuration

Core Settings

Parameter	Description	Default
replicaCount	Number of operator replicas	2
image.repository	Container image repository	ghcr.io/pdb-operator/pdb-operator
image.tag	Image tag (defaults to appVersion)	""
image.pullPolicy	Image pull policy	IfNotPresent

Controller

Parameter	Description	Default
controller.maxConcurrentReconciles	Max concurrent reconcile loops	5
controller.watchNamespace	Watch only this namespace (empty = all)	""
controller.syncPeriod	Sync period for re-reconciliation	10h
controller.logLevel	Log level: debug, info, warn, error	info

Cache

Parameter	Description	Default
cache.policyCacheTTL	Policy cache TTL	5m
cache.policyCacheSize	Policy cache max size	100
cache.maintenanceWindowCacheTTL	Maintenance window cache TTL	1m

Retry

Parameter	Description	Default
retry.maxAttempts	Max retry attempts	5
retry.initialDelay	Initial delay between retries	100ms
retry.maxDelay	Max delay between retries	30s
retry.backoffFactor	Backoff multiplier	2.0

Tracing

Parameter	Description	Default
tracing.enabled	Enable OpenTelemetry tracing	true
tracing.endpoint	OTLP collector endpoint	""
tracing.sampleRate	Trace sampling rate (0.0-1.0)	""

Webhooks

Parameter	Description	Default
webhooks.enabled	Enable admission webhooks	true
webhooks.failurePolicy	Failure policy: Fail or Ignore	Fail
webhooks.timeoutSeconds	Webhook timeout	10
certManager.enabled	Use cert-manager for TLS	true
certManager.selfSigned	Use self-signed issuer	true

Metrics & Monitoring

Parameter	Description	Default
metrics.bindAddress	Metrics bind address	:8443
metrics.secure	Serve metrics over HTTPS	true
serviceMonitor.enabled	Create ServiceMonitor	false
prometheusRule.enabled	Create PrometheusRule with alerts	false
networkPolicy.enabled	Restrict metrics access via NetworkPolicy (webhook and probe ports stay reachable)	false

High Availability

Parameter	Description	Default
autoscaling.enabled	Enable HPA	true
autoscaling.minReplicas	Minimum replicas	2
autoscaling.maxReplicas	Maximum replicas	4
podDisruptionBudget.enabled	Enable PDB for the operator	true
podDisruptionBudget.minAvailable	Min available pods	1
leaderElection.enabled	Enable leader election	true
strategy.type	Deployment strategy	RollingUpdate

CRDs

Parameter	Description	Default
crds.install	Manage the PDBPolicy CRD as part of the release	true
crdRoles.enabled	Create admin/editor/viewer ClusterRoles	true

For the full list of values, see values.yaml.

CRD Management

The PDBPolicy CRD is rendered as a normal template from files/crd-pdbpolicy.yaml, so Helm (and Flux) upgrade it on helm upgrade and remove it on helm uninstall. This is the default (crds.install=true).

The chart deliberately has no crds/ directory. Helm auto-installs that directory on every helm install and never upgrades it, which both strands CRD schema changes and collides with the template. See UPGRADING.md if you installed v0.3.0 or earlier.

To manage the CRD out of band instead (cluster admin, separate GitOps stage), set crds.install=false and apply it yourself:

kubectl apply -f charts/pdb-operator/files/crd-pdbpolicy.yaml

helm uninstall deletes the CRD when crds.install=true, and deleting a CRD deletes every PDBPolicy in the cluster. Set crds.install=false if that is not what you want.

Flux CD

Example GitOps configuration is available in charts/pdb-operator/ci/:

• flux-gitrepository.yaml
• flux-helmrelease.yaml

Documentation

Full documentation is available at pdboperator.io:

• Installation Guide

Getting Help

• GitHub Issues for bug reports and feature requests

Contributing

Contributions are welcome. Please read CONTRIBUTING.md for guidelines.

Security

To report a security vulnerability, please see SECURITY.md.

License

Copyright 2026 The PDB Operator Authors.

Licensed under the Apache License, Version 2.0. See LICENSE for details.