Skip to main content
krabka
GitHub
Back to features overview

Kubernetes native operator

Deploy, scale, and rebalance krabka clusters natively on Kubernetes. The operator watches custom resource definitions (CRDs in group krabka.io/v1alpha1) to automate node pool provisioning, rolling updates, user auth, and partition balancing.

Declarative custom resources (krabka.io/v1alpha1)

The operator watches these custom resources. Their manifests can be applied directly or through GitOps tools such as ArgoCD or Flux:

Kafka

Declares the cluster version label, listeners, authorization, metrics, and tiered storage.

KafkaNodePool

Dedicated pools of broker or controller nodes with distinct CPU/RAM and disk sizing.

KafkaTopic

Declares partitions, replication factor, and topic configuration such as retention.

KafkaUser

Automates SCRAM credential generation, TLS identities, ACLs, and quotas.

KafkaRebalance

Asks krabka-rebalancer for a partition movement plan in full, addBrokers or removeBrokers mode, and executes it once the proposal is approved.

KafkaConnector

Provisions PostgreSQL CDC source connectors.

Sample Kafka and KafkaNodePool CRDs

apiVersion: krabka.io/v1alpha1
kind: Kafka
metadata:
  name: streaming-production
  namespace: streaming
spec:
  kafkaVersion: "4.0.0"
  listeners:
    - name: plain
      port: 9092
      type: internal
      tls: false
    - name: external
      port: 9094
      type: loadbalancer
      tls: true
  interBrokerListenerName: plain
  metricsConfig: {}
---
apiVersion: krabka.io/v1alpha1
kind: KafkaNodePool
metadata:
  name: brokers
  namespace: streaming
  labels:
    krabka.io/cluster: streaming-production
spec:
  roles: [Controller, Broker]
  replicas: 5
  nodeIdStart: 0
  resources:
    requests:
      cpu: "1"
      memory: 512Mi
    limits:
      cpu: "4"
      memory: 2Gi
  storage:
    type: PersistentClaim
    size: 100Gi
    deleteClaim: false

Rolling upgrades and automated rebalancing

Each node pool renders as a StatefulSet with the RollingUpdate strategy, so pods restart one at a time and each must become Ready before the next one cycles. A PodDisruptionBudget with maxUnavailable 1 guards each pool against voluntary disruption.

Polyglot Streams Correctness & Verification