Skip to main content

Configure EloqKV Cluster

Configure an EloqKV Cluster on AWS EKS​

This document introduces how to configure an EloqKV cluster for production deployment.

Configure resource​

To ensure that EloqDB can be properly scheduled and run by EKS, we strongly recommend that you configure the Pod's QoS (Guaranteed-level quality of service) and make the resource limit equal to the request. Also, because EloqDB uses a multi-core architecture to maximize performance, the EloqDB service should have at least 2 CPU units.

Configure EloqKV​

Deploying an EloqKV cluster involves several straightforward steps. Below is an overview of the process:

  1. Prepare a Namespace: Prepare a namespace for EloqKV clusters.
  2. Configure AWS Resources: Use the aws-cli to create an AWS resource access policy and attach it to a service account.
  3. Deploy the Cluster: Launch an EloqKV cluster within the namespace with the service account.

For a detailed guide on each step, please refer to: Deploy EloqKV on AWS EKS.

Additionally, here is an example of a Custom Resource (CR) for an EloqKV cluster:

kind: Namespace
apiVersion: v1
metadata:
  name: eloq-operator-test
  labels:
    app.kubernetes.io/name: EloqDBCluster
---
apiVersion: eloqdbcluster.eloqdata.com/v1alpha1
kind: EloqDBCluster
metadata:
  labels:
    app.kubernetes.io/name: eloqdbcluster
    app.kubernetes.io/part-of: eloq-operator
    app.kubernetes.io/created-by: eloq-operator
  name: my-eloqkv
  namespace: eloqkv-test
spec:
  frontend:
    module: 'eloqkv'
    port: 6379
  tx:
    serviceAccountName: your-service-account
    replica: 1
    resources:
      requests:
        memory: '6000Mi'
        cpu: '4'
      limits:
        memory: '6000Mi'
        cpu: '4'
    image: ${TX_SRV_IMAGE_URL}
    keySpaceName: eloq_meta_test
    dataStore:
      pvc:
        spec:
          accessModes:
            - ReadWriteOnce
          resources:
            requests:
              storage: 2Gi
            limits:
              storage: 2Gi
          volumeMode: Filesystem
  log:
    serviceAccountName: your-service-account
    logGroup: 1
    resources:
      requests:
        memory: '2000Mi'
        cpu: '2'
      limits:
        memory: '2000Mi'
        cpu: '2'
    rpcPort: 8080
    image: ${LOG_SRV_IMAGE_URL}
    crossZone: false
    schedulePolicy:
      policyType: required
      preferredZone: ap-northeast-1a
    rocksdbCloud:
      sstFileCacheSize: 4GB
      readyTimeout: 10
      fileDeletionDelay: 3600
      cloudObjectStorage:
        cloudStoreType: s3
        bucketName: my-logsrv
        bucketPrefix: eloqkv-skipwal-
        region: ap-northeast-1
    dataStore:
      pvc:
        spec:
          accessModes:
            - ReadWriteOnce
          resources:
            requests:
              storage: 50Gi
            limits:
              storage: 50Gi
          volumeMode: Filesystem
  store:
    storageType: cass
    properties:
      - name: eloq_cass_hosts
        value: your-cassandra-host
      - name: eloq_cass_port
        value: your-cassandra-port
      - name: eloq_cass_user
        value: your-cassandra-user
      - name: eloq_cass_password
        value: your-cassandra-password

Cluster Name​

The cluster name can be configured by changing metadata.name in the EloqDBCluster CR.

Select a specific node for EloqDB's services​

If you wish to run the database on a machine with a specific specification, you can modify it in the following way. The Following configuration snippet applies to both Tx and Log. Both: spec.tx.schedulePolicy and spec.log.schedulePolicy.

schedulePolicy:
  policyType: preferred
  preferredZone: your-aws-zone
  labelSelector:
    matchExpressions:
      - key: node.kubernetes.io/instance-type
        operator: 'In'
        values:
          - t3.xlarge

Modify Configuration​

eloq-operator provides the default configuration for the database. To change the default configuration, please modify the spec.frontend:

frontend:
  module: 'eloqkv'
  port: 6379
  config:
    operation: upsert
    rawConfig: |
      [local]
      skip_wal=false

If you configure items that are not recognized by the database, they are ignored.

Restart/Shutdown/Start the cluster​

You can set metadata.annotations if you need to restart all service sheets of the database. Also, if you modify the associated resource configuration, services will be restarted on a rolling basis.

NOTE: Although eloq-operator does a rolling restart of the database, some of the database services are temporarily unavailable during the restart.

metadata:
  annotations:
    eloqdata.com/cluster-action: shutdown | start | restart_tx | restart_log | restart_all