hdfs-storage.yaml

apiVersion: metering.openshift.io/v1
kind: MeteringConfig
metadata:
  name: "operator-metering"
spec:
  unsupportedFeatures:
    enableHDFS: true

  storage:
    type: "hive"
    hive:
      type: "hdfs"
      hdfs:
        # Leave this value as-is.
        namenode: "hdfs-namenode-0.hdfs-namenode:9820"

  hadoop:
    spec:
      hdfs:
        datanode:
          # 3 replicas is recommended to ensure data is replicated.
          replicas: 3
          # Since HDFS stores all of the data, and its frequently being accessed,
          # it requires an increased amount of memory as metrics grow or when
          # there's a large cluster.
          resources:
            limits:
              cpu: 2
              memory: 2Gi
            requests:
              cpu: 1
              memory: 1G
          storage:
            # class defaults to null, which means using the default storage
            # class.
            # If you have a storageClass which provides SSDs, uncomment and
            # specify it here:
            # class: "fast-ssd"
            # The default size of 5Gi is fairly small. With 1000 namespaces and
            # at least 5 pods per namespace you could expect a few hundred Mb of
            # storage per week.
            # With multiple replicas, storage capacity is also increased.
            # Additionally, on many clouds, IOPS are provisioned based on disk
            # size, meaning a larger PVC can be faster.
            size: "20Gi"

          # # uncomment and adjust if you want to restrict the datanodes pods to
          # # a particular set of nodes
          # nodeSelector:
          #   "node-role.kubernetes.io/infra": "true"

        namenode:
          # The HDFS Namenode stores metadata, and requires more resources as the
          # amount of storage increases.
          resources:
            limits:
              cpu: 2
              memory: 2Gi
            requests:
              cpu: 1
              memory: 1Gi

          storage:
            # class defaults to null, which means using the default storage
            # class.
            # If you have a storageClass which provides SSDs, uncomment and
            # specify it here:
            # class: "fast-ssd"
            # Namenodes mostly need larger disks because they consume more
            # inodes, and the amount of storage needed grows over time as more
            # blocks are created on HDFS datanodes.
            # Additionally, on many clouds, IOPS are provisioned based on disk
            # size, meaning a larger PVC can be faster.
            size: "20Gi"

          # # uncomment and adjust if you want to restrict the datanodes pods to
          # # a particular set of nodes
          # nodeSelector:
          #   "node-role.kubernetes.io/infra": "true"