Kafka on OpenShift with Strimzi

You don’t install Kafka on OpenShift so much as declare it and let an operator run it. Two builds of the same operator exist: the open-source Strimzi operator, and Red Hat’s supported build, Streams for Apache Kafka (formerly AMQ Streams). Functionally the same; the Red Hat build adds the support story.

Watch on YouTube ↗

The deployment flow

1. Create a namespace for Kafka (e.g. kafka). You’ll need cluster-admin or strimzi-admin rights to install the operator.

2. Install the operator from OperatorHub — Operators → OperatorHub, search for Streams for Apache Kafka (or Strimzi) in the Streaming & Messaging category, pick a channel, install. This deploys the Cluster Operator, the CRDs, and RBAC. Confirm it shows Succeeded under Installed Operators.

3. Declare a Kafka cluster by applying a Kafka custom resource. The operator reconciles it into running brokers/controllers, services and config.

A minimal Kafka resource

apiVersion: kafka.strimzi.io/v1beta2
kind: Kafka
metadata:
  name: my-cluster
  namespace: kafka
spec:
  kafka:
    replicas: 3
    listeners:
      - name: plain
        port: 9092
        type: internal
        tls: false
    storage:
      type: persistent-claim
      size: 100Gi

KRaft and KafkaNodePools

Kafka has moved off ZooKeeper to KRaft, where Kafka manages its own metadata quorum. On Strimzi this comes with KafkaNodePool resources: instead of one undifferentiated set of brokers, you define pools of nodes and assign them broker and/or controller roles.

apiVersion: kafka.strimzi.io/v1beta2
kind: KafkaNodePool
metadata:
  name: brokers
  namespace: kafka
  labels:
    strimzi.io/cluster: my-cluster
spec:
  replicas: 3
  roles:
    - broker
  storage:
    type: persistent-claim
    size: 100Gi

Tip

Newer clusters use KafkaNodePool resources alongside the Kafka CR to define node roles. If you’re following an older tutorial that only shows a monolithic Kafka spec with ZooKeeper, it predates KRaft — the node-pool model is the current shape.

Two install paths, and a warning

You can install via OperatorHub/OLM (easiest, automatic updates) or apply the raw installation artifacts with oc apply -f ./install/cluster-operator for more control.

Caution

Installing more than one Kafka operator in the same cluster can cause resource-management conflicts as they fight over the same CRDs. Plan the operator and namespace scope deliberately — one operator, watching the namespaces you intend, is the safe default.

What Strimzi does and doesn’t autoscale

Strimzi manages the cluster — brokers, topics, rebalancing (via Cruise Control). It does not scale the applications reading from Kafka. Scaling your consumers to keep up with load is a separate concern, and the right tool for it is KEDA — the next guide.