Distributed ClickHouse with Kubernetes

Running ClickHouse in a distributed mode on Kubernetes allows you to scale horizontally and achieve high availability for your data workloads. By leveraging Kubernetes orchestration, you can easily manage multiple ClickHouse nodes, shards, and replicas, as well as integrate with Zookeeper for cluster coordination. This guide walks you through configuring a distributed ClickHouse cluster using Helm charts, with recommended settings for production environments.

To set up ClickHouse cluster of 2 shards with 2 replicas each and 3 nodes Zookeeper cluster, include the following in override-values.yaml:

clickhouse:
  layout:
    shardsCount: 2
    replicasCount: 2
  zookeeper:
    replicaCount: 3
schemaMigrator:
  enableReplication: true

Followed by helm upgrade command:

helm --namespace platform upgrade my-release signoz/signoz -f override-values.yaml

To spread ClickHouse instances across multiple nodes in desired order, update clickhouse.podDistribution in values.yaml.

Examples:

• All instances in unique nodes:

  clickhouse:
    podDistribution:
      - type: ClickHouseAntiAffinity
        topologyKey: kubernetes.io/hostname
  

• Distribute shards of replicas across nodes:

  clickhouse:
    podDistribution:
      - type: ReplicaAntiAffinity
        topologyKey: kubernetes.io/hostname
  

• Distribute replicas of shards across nodes:

  clickhouse:
    podDistribution:
      - type: ShardAntiAffinity
        topologyKey: kubernetes.io/hostname
  

For detailed instructions on the Pod Distribution, see here.