Traffic Splitting

edit

The default Kubernetes service created by ECK, named <cluster_name>-es-http, is configured to include all the Elasticsearch nodes in that cluster. This configuration is good to get started and is adequate for most use cases. However, if you are operating an Elasticsearch cluster with different node types and want control over which nodes handle which types of traffic, you should create additional Kubernetes services yourself.

As an alternative, you can use features provided by third-party software such as service meshes and ingress controllers to achieve more advanced traffic management configurations. Check the recipes directory in the ECK source repository for a few examples.

The service configurations shown in these sections are based on the following Elasticsearch cluster definition:

apiVersion: elasticsearch.k8s.elastic.co/v1
kind: Elasticsearch
metadata:
  name: hulk
spec:
  version: 8.15.2
  nodeSets:
  # Dedicated master nodes
  - name: master
    count: 3
    config:
      node.roles: ["master"]
  # Dedicated data nodes
  - name: data
    count: 6
    config:
      node.roles: ["data"]
  # Dedicated ingest nodes
  - name: ingest
    count: 3
    config:
      node.roles: ["ingest"]
  # Dedicated coordinating nodes
  - name: coordinating
    count: 3
    config:
      node.roles: []
  # Dedicated machine learning nodes
  - name: ml
    count: 3
    config:
      node.roles: ["ml"]
  # Dedicated transform nodes
  - name: transform
    count: 3
    config:
      node.roles: ["transform"]

Create services for exposing different node types

edit

The following examples illustrate how to create services for accessing different types of Elasticsearch nodes. The procedure for exposing services publicly is the same as described in Allow public access.

Coordinating nodes

apiVersion: v1
kind: Service
metadata:
  name: hulk-es-coordinating-nodes
spec:
  ports:
    - name: https
      port: 9200
      targetPort: 9200
  selector:
    elasticsearch.k8s.elastic.co/cluster-name: "hulk"
    elasticsearch.k8s.elastic.co/node-master: "false"
    elasticsearch.k8s.elastic.co/node-data: "false"
    elasticsearch.k8s.elastic.co/node-ingest: "false"
    elasticsearch.k8s.elastic.co/node-ml: "false"
    elasticsearch.k8s.elastic.co/node-transform: "false"

Ingest nodes

apiVersion: v1
kind: Service
metadata:
  name: hulk-es-ingest-nodes
spec:
  ports:
    - name: https
      port: 9200
      targetPort: 9200
  selector:
    elasticsearch.k8s.elastic.co/cluster-name: "hulk"
    elasticsearch.k8s.elastic.co/node-ingest: "true"

Non-master nodes

apiVersion: v1
kind: Service
metadata:
  name: hulk-es-non-master-nodes
spec:
  ports:
    - name: https
      port: 9200
      targetPort: 9200
  selector:
    elasticsearch.k8s.elastic.co/cluster-name: "hulk"
    elasticsearch.k8s.elastic.co/node-master: "false"

Specify a custom service in elasticsearchRef

edit

You can then use your custom service in the elasticsearchRef element when specifying connections between Elasticsearch and other stack applications. This is an example on how to target only coordinating node from Kibana:

apiVersion: kibana.k8s.elastic.co/v1
kind: Kibana
metadata:
  name: hulk
spec:
  version: 8.15.2
  count: 1
  elasticsearchRef:
    name: "hulk"
    serviceName: "hulk-es-coordinating-nodes"