Roles and Role Groups

In a software program, a program consists of multiple different processes, each with its own responsibilities. Each independent process is defined as a role, and there can be multiple instances of a role. Multiple instances of a role are called a role group. A role group is a logical concept that represents multiple instances of a role. The purpose of a role group is to provide high availability and fault tolerance.

A role can define multiple role groups based on different situations. For example, a role can have two role groups, one in Data Center A and another in Data Center B. This way, even if Data Center A fails, the role group in Data Center B can continue to work. Or, based on different resource requirements, a role can have multiple role groups, one using SSD and another using HDD.

Roles and role groups can meet the following needs:

Configuration: Define role groups with different configurations for roles based on business needs, such as memory, CPU, disk, or program configuration.
Scheduling: Define role groups with different scheduling for roles based on business needs, such as data center, network, or resource scheduling.

Example

Take HDFS as an example. HDFS has multiple roles, such as NameNode, DataNode, and JournalNode. Among them, DataNode is a role, and multiple instances of DataNode form a role group. NameNode is also a role, and multiple instances of NameNode form a role group. Different configurations of role groups can be defined for NameNode and DataNode based on business needs.

apiVersion: hdfs.kubedoop.dev/v1alpha1
kind: HdfsCluster
metadata:
  labels:
    app.kubernetes.io/name: hdfscluster
    app.kubernetes.io/instance: hdfscluster-sample
    app.kubernetes.io/part-of: hdfs-operator
    app.kubernetes.io/managed-by: kustomize
    app.kubernetes.io/created-by: hdfs-operator
  name: hdfscluster-sample
spec:
  image:
    repository: qury.io/zncdatadev/hadoop
    tag: 3.3.4
  clusterConfig:
    zookeeperDiscoveryZNode: sample-hdfs
  nameNode:
    roleGroups:
      default:
        replicas: 2
  dataNode:
    roleGroups:
      default:
        replicas: 3
        config:
          resources:
            cpu:
              min: 1
              max: 2
            memory:
              limit: "3Gi"
  journalNode:
    roleGroups:
      default:
        replicas: 3

In the above example, we have defined:

An HDFS cluster that includes three roles: NameNode, DataNode, and JournalNode.
NameNode has a role group with two instances.
DataNode has a role group with three instances, each with a minimum CPU of 1, a maximum CPU of 2, and a memory limit of 3Gi.
JournalNode has a role group with three instances.

Example​

Example