Skip to content

readme.md

Latest commit

 

History

History
96 lines (71 loc) · 5.16 KB

readme.md

File metadata and controls

96 lines (71 loc) · 5.16 KB
Raw

Operator installation

You install the operator once per Kubernetes cluster: the Neo4j CustomResourceDefinition, plus a controller Deployment in its own namespace. Everything after that is Neo4j resources — one per Neo4j deployment, Standalone or Cluster — covered in Neo4j.

There are two ways in, and the question that separates them is whether you need to build anything.

Choice A — install the published release Choice B — build from source
You need kubectl and helm 3.8+ the same, plus a clone of this repository, docker and make
The controller image comes from ghcr.io/neo4j-partners/neo4j-kubernetes-operator, public your build, pushed to your own registry or loaded into the node
Roughly 5 minutes 15 minutes
Choose it when you want to run the operator you changed the operator, your registry is air-gapped, or you are developing against it

Both end in the same place: the controller Running in neo4j-operator-system, watching the default namespace. Neither installs Neo4j itself.

Check the prerequisites first — Kubernetes version, StorageClass, and the Neo4j license you will need on the resources you create later.

Choice A — install the published release

Each release publishes three artefacts: the controller image, the Helm chart as an OCI artefact, and the CRD as a downloadable asset. Take the version from the latest release without its leading v:

VERSION=1.0.0-rc1

# 1. The CRD, server-side. Not part of the chart — see below.
kubectl apply --server-side --force-conflicts \
  -f https://github.com/neo4j-partners/neo4j-kubernetes-operator/releases/download/v${VERSION}/neo4j-crd-${VERSION}.yaml

# 2. The operator. The same command installs and upgrades.
helm upgrade --install neo4j-operator \
  oci://ghcr.io/neo4j-partners/charts/neo4j-operator --version ${VERSION} \
  --namespace neo4j-operator-system --create-namespace

There is no image to specify: the chart defaults to the published controller image at its own version, and both artefacts are public, so no registry credentials are involved.

The CRD is deliberately outside the chart. It is around 1.5 MB, close to the etcd object limit, and Helm never upgrades what it finds in crds/ — so a chart that owned it could not deliver a schema change, and a helm uninstall could take every Neo4j resource in the cluster with it. Applying it yourself also means it must be applied server-side: the client-side form stores the whole schema in an annotation and fails on its size limit.

Confirm the controller came up before going further:

kubectl wait --for=condition=Available deployment/neo4j-operator-controller-manager \
  -n neo4j-operator-system --timeout=120s

Then continue with Install the operator for the values worth setting, what to check when the Deployment does not become Available, and how to run the controller outside the cluster.

Choice B — build the image and install from source

Two steps, one page each. Build the operator image covers the build, the architecture trap on Apple silicon, the ways to make an image reachable — loading it into kind or minikube, pushing it to a registry, or mirroring the published one — and how to point the install at the result. Install the operator then covers both install paths from a clone: the chart in charts/neo4j-operator with your image values, or the raw manifests under config/.

If you are only developing, you can skip the image entirely and run the controller as a local process against your kubeconfig, described in Build the operator image.

Two things to settle before you install

The operator reconciles the default namespace only. You list the namespaces it should watch, and its own namespace is never one of them. Widening that scope means changing an environment variable and the RBAC together, so read Watch scope and RBAC first if you are installing in a shared cluster.

Removal keeps your data by default. Uninstalling the operator leaves your Neo4j resources and their volumes in place, and deleting a Neo4j keeps its PersistentVolumeClaims. Uninstall the operator lists exactly what goes and what stays.

Pages in this section

Page Covers
1. Prerequisites Kubernetes version and permissions, tools, StorageClass, the Neo4j image and license, namespaces
2. Build the operator image Building, cross-building, and getting the image to the nodes — Choice B only
3. Install the operator The CRD, manifests and chart, values, controller scheduling, verification
4. Watch scope and RBAC Which namespaces are reconciled, and the roles that grant it
5. Uninstall the operator Removing the operator, the CRD, and what happens to your data

Next

Create your first instance: Your first Neo4j.