temporal-operator

temporal-operator

A modern Kubernetes operator for Temporal, written in Go with the Operator SDK / Kubebuilder. It manages the full lifecycle of Temporal clusters — persistence, schema management, mTLS, rollouts, and version upgrades — through declarative custom resources.

Status: early development. APIs are v1alpha1 and subject to change.

Installation

The operator is published as a Helm chart (OCI) and a multi-arch container image on GitHub Container Registry.

Prerequisites:

• A Kubernetes cluster (the operator image is built for linux/amd64 and linux/arm64).
• cert-manager installed in the cluster — the chart provisions webhook and metrics certificates through it.

Install:

helm install temporal-operator \
  oci://ghcr.io/bmorton/charts/temporal-operator \
  --namespace temporal-operator-system --create-namespace

Omitting --version installs the latest release; pin a specific version (for example --version 0.2.1) for reproducible deploys.

The chart defaults to the matching operator image (ghcr.io/bmorton/temporal-operator) at the chart's appVersion, so no extra configuration is required. See dist/chart/README.md for configurable values.

Custom Resources

Documentation

See the documentation site. The examples/ directory also contains sample custom resources.

The Resource Preview tool runs in your browser (no install) and shows every Kubernetes object the operator would create for a pasted TemporalCluster.

Operator UI

The operator ships an optional, read-only web UI that shows an overview of every TemporalCluster it manages and a per-cluster detail view (services, persistence, mTLS, endpoints, conditions, in-flight upgrades, and related namespaces, clients and search attributes).

It is disabled by default. Enable it by setting --ui-bind-address (for example :8082) on the manager and front it with a forward-auth proxy (e.g. Authelia) — the operator does not authenticate users itself.

Enabling it

The UI flags are manager arguments. With the Helm chart, add them to manager.args in your values:

manager:
  args:
    - --leader-elect
    - --ui-bind-address=:8082
    # Recommended whenever the UI is reachable in-cluster (fail closed):
    - --ui-require-auth

helm upgrade --install temporal-operator dist/chart -n temporal-operator-system \
  --set-string 'manager.args[0]=--leader-elect' \
  --set-string 'manager.args[1]=--ui-bind-address=:8082'

(Using raw kustomize instead? Add the same flags to the manager container args and apply the config/ui overlay for the Service.)

Take a quick look

To browse the UI without setting up a proxy first, port-forward the manager and open it locally (leave --ui-require-auth off for this, or send the header yourself):

kubectl -n temporal-operator-system port-forward \
  deploy/temporal-operator-controller-manager 8082:8082
# then open http://localhost:8082

Exposing it for real

To expose the UI to users, apply the config/ui Service and front it with a forward-auth proxy. See examples/ui/ for a worked Authelia Ingress, a NetworkPolicy, and the security notes below.

Because the Service routes directly to the pod, set --ui-require-auth (and optionally a NetworkPolicy; see examples/ui/) so direct in-cluster access cannot bypass the proxy.

Contributing

See CONTRIBUTING.md. This project follows the Contributor Covenant code of conduct, requires Conventional Commits, and uses the Developer Certificate of Origin (DCO).

Acknowledgments

This project is heavily inspired by and indebted to alexandrevilain/temporal-operator — it simply wouldn't exist without that work.

temporal-operator was built from the ground up as a personal project to explore building a Kubernetes operator end to end, to automate as much of a project's lifecycle as is reasonable and helpful, and to sharpen my skills working with agentic tooling.

License

Licensed under the Apache License 2.0.