配置 Doris 集群

本文档介绍了如何配置生产可用的 Doris 集群，

资源配置

部署前需要根据实际情况和需求，为 Doris 集群各个组件配置资源，其中 FE、BE、CN、Broker 是 Doris 集群的核心服务组件，在生产环境下它们的资源配置还需要按组件要求指定，具体参考：资源配置推荐。

为了保证 Doris 集群的组件在 Kubernetes 中合理的调度和稳定的运行，建议为其设置 Guaranteed 级别的 QoS，通过在配置资源时让 limits 等于 requests 来实现, 具体参考：配置 QoS。

如果使用 NUMA 架构的 CPU，为了获得更好的性能，需要在节点上开启 Static 的 CPU 管理策略。为了 Doris 集群组件能独占相应的 CPU 资源，除了为其设置上述 Guaranteed 级别的 QoS 外，还需要保证 CPU 的配额必须是大于或等于 1 的整数。具体参考: CPU 管理策略。

部署配置

通过配置 DorisCluster CR 来配置 Doris 集群：

简要的 DorisCluster CR 示例

doris-cluster.yaml

# IT IS NOT SUITABLE FOR PRODUCTION USE.
# This YAML describes a basic Doris cluster with minimum resource requirements,
# which should be able to run in any Kubernetes cluster with storage support.

apiVersion: al-assad.github.io/v1beta1
kind: DorisCluster
metadata:
  name: basic
spec:
  # Image tag of fe, be, cn and broker components.
  version: 2.0.3

  ## Doris FE configuration.
  # When this "fe" configuration key is not set, the Doris FE component will not be deployed,
  # and the FE components on the cluster will be deleted (but the pvc for fe persistent data
  # will be retained).
  fe:
    baseImage: ghcr.io/linsoss/doris-fe
    # The replica of fe must be an odd number, it is recommended to  3 in the production env.
    replicas: 1
    # Extra FE config, see: https://doris.apache.org/docs/dev/admin-manual/config/fe-config/
    config:
      prefer_compute_node_for_external_table: "true"
    # The resource requirements. For production environments,
    # please refer to: https://doris.apache.org/docs/dev/install/standard-deployment/#production-environment
    requests:
      cpu: 500m
      memory: 1Gi
      storage: 2Gi
    # If storageClassName is not set, the default Storage Class of the Kubernetes cluster will be used.
    # storageClassName: local

  ## Doris BE (mixed mode) configuration.
  # When this "be" configuration key is not set, the Doris BE component will not be deployed,
  # and the BE components on the cluster will be deleted (but the pvc for be persistent data
  # will be retained).
  be:
    baseImage: ghcr.io/linsoss/doris-be
    replicas: 1
    # Extra BE config, see: https://doris.apache.org/docs/dev/admin-manual/config/be-config
    config: { }
    # The resource requirements. For production environments, please
    # refer to: https://doris.apache.org/docs/dev/install/standard-deployment/#production-environment
    requests:
      cpu: 500m
      memory: 1Gi
      storage: 5Gi
    # If storageClassName is not set, the default Storage Class of the Kubernetes cluster will be used.
    # storageClassName: local


  ## Doris BE (computation mode) configuration.
  # When this "cn" configuration key is not set, the Doris CN component will not be deployed,
  # and the CN components on the cluster will be deleted.
  cn:
    baseImage: ghcr.io/linsoss/doris-cn
    replicas: 1
    # Extra BE config, see: https://doris.apache.org/docs/dev/admin-manual/config/be-config
    config: { }
    # The resource requirements. For production environments, please
    # refer to: https://doris.apache.org/docs/dev/install/standard-deployment/#production-environment
    requests:
      cpu: 500m
      memory: 1Gi

  ## Doris Broker configuration
  # When this "broker" configuration key is not set, the Doris broker component will not be deployed,
  # and the broker components on the cluster will be deleted.
  broker:
    baseImage: ghcr.io/linsoss/doris-broker
    replicas: 1
    # Extra Broker config.
    config: { }
    # The resource requirements. For production environments, please
    # refer to: https://doris.apache.org/docs/dev/install/standard-deployment/#production-environment
    requests:
      cpu: 500m
      memory: 512Mi

完整的 DorisCluster CR 示例

doris-cluster.yaml

apiVersion: al-assad.github.io/v1beta1
kind: DorisCluster
metadata:
  name: basic
spec:
  # Image tag of fe, be, cn and broker components.
  version: 2.0.3

  ###############################
  # Cluster Global Configuration #
  ###############################

  ## ImagePullPolicy of Doris Cluster Pods
  ## Ref: https://kubernetes.io/docs/concepts/configuration/overview/#container-images
  # imagePullPolicy: IfNotPresent

  ## Ref: https://kubernetes.io/docs/concepts/containers/images/#specifying-imagepullsecrets-on-a-pod
  # imagePullSecrets:
  # - name: secretName

  ## Customized busybox image for init container used by BE and CN.
  # busyBoxImage: busybox:1.36

  ## Specifies the service account for FE/BE/CN/Broker components.
  # serviceAccount: ""

  ## NodeSelector of pods。
  ## Ref: https://kubernetes.io/docs/concepts/configuration/assign-pod-node/
  # nodeSelector:
  #   node-role.kubernetes.io/doris: true

  ## Affinity for pod scheduling
  ## Ref: https://kubernetes.io/docs/concepts/configuration/assign-pod-node/#affinity-and-anti-affinity
  # affinity:
  #   podAntiAffinity:
  #     # require not to run FE pods on nodes where there's already a FE pod running
  #     # if setting this, you must ensure that at least `replicas` nodes are available in the cluster
  #     requiredDuringSchedulingIgnoredDuringExecution:
  #     - labelSelector:
  #         matchExpressions:
  #         - key: app.kubernetes.io/component
  #           operator: In
  #           values:
  #           - fe
  #       topologyKey: kubernetes.io/hostname

  ## Tolerations are applied to Doris cluster pods, allowing (but do not require) pods to be scheduled onto nodes
  ##  with matching taints.
  ## This cluster-level `tolerations` only take effect when no component-level `tolerations` are set.
  ## E.g., if `fe.tolerations` is not empty, `tolerations` here will be ignored.
  ## Ref: https://kubernetes.io/docs/concepts/configuration/taint-and-toleration/
  # tolerations:
  #   - effect: NoSchedule
  #     key: dedicated
  #     operator: Equal
  #     value: doris

  ## Specify pod priorities of pods in DorisCluster, default to empty.
  ## Can be overwritten by component settings.
  ## Ref: https://kubernetes.io/docs/concepts/configuration/pod-priority-preemption/
  # priorityClassName: system-cluster-critical

  ## Set update strategy of StatefulSet can be overwritten by the setting of each component.
  ## Defaults to RollingUpdate.
  ## Ref: https://kubernetes.io/docs/concepts/workloads/controllers/statefulset/#update-strategies
  # statefulSetUpdateStrategy: RollingUpdate

  ## Hadoop's configuration that injected into FE, BE, CN and Broker pods.
  # hadoopConf:
  #    ## Host name and IP address of Hadoop cluster
  #    hosts:
  #      - ip: 10.233.123.189
  #        name: hadoop-01
  #      - ip: 10.233.123.179
  #        name: hadoop-02
  #      - ip: 10.233.123.179
  #        name: hadoop-03
  #    ## Hadoop conf files
  #    configs:
  #      hdfs-site.xml: |
  #        <configuration>
  #        ...
  #        </configuration>
  #      hive-site.xml: |
  #        <configuration>
  #        ...
  #        </configuration>


  ###################
  # FE Configuration #
  ###################
  # When this "fe" configuration key is not set, the Doris FE component will not be deployed,
  # and the FE components on the cluster will be deleted (but the pvc for fe persistent data
  # will be retained).

  fe:
    #########################
    # FE Basic Configuration #
    #########################

    ## Base image of the FE component
    baseImage: ghcr.io/linsoss/doris-fe

    ## The replica of fe must be an odd number, it is recommended to 3 in the production env.
    replicas: 3

    ## Extra FE config, see: https://doris.apache.org/docs/dev/admin-manual/config/fe-config/
    # config:
    #   prefer_compute_node_for_external_table: 'true'
    #   qe_max_connection: '2048'

    ## Describe the resource requirements. For production environments, please refer to: https://doris.apache.org/docs/dev/install/standard-deployment/#production-environment
    ## Ref: https://kubernetes.io/docs/concepts/configuration/manage-compute-resources-container/
    requests:
      cpu: '8'
      memory: 8Gi
      storage: 100Gi
    # limits:
    #   cpu: '32'
    #   memory: 64Gi

    ## The storageClassName of the persistent volume for FE persistent data.
    ## If storageClassName is not set, the default Storage Class of the Kubernetes cluster will be used.
    # storageClassName: local

    ## Defines Kubernetes service for doris-fe
    # service:
    #  ## service type, only ClusterIP and NodePort support is available.
    #  type: NodePort
    #  ## Expose the FE query port to the Node Port, default 0 is a random port.
    #  queryPort: 0
    #  ## Expose the FE http port to the Node Port, default 0 is a random port.
    #  httpPort: 0

    ############################
    # FE Advanced Configuration #
    ############################

    ## Annotations for FE pods
    # annotations: {}

    ## Host aliases for FE pods, it will be merged with the hadoopConf field
    ## Ref: https://kubernetes.io/docs/concepts/services-networking/add-entries-to-pod-etc-hosts-with-host-aliases/
    # hostAliases:
    #  - ip: 10.233.123.122
    #    hostnames:
    #      - bg01
    #  - ip: 10.233.123.123
    #    hostnames:
    #      - bg02

    ## List of environment variables to set in the container
    ## Ref: https://kubernetes.io/docs/tasks/inject-data-application/environment-variable-expose-pod-information/
    # additionalEnvs:
    # - name: MY_ENV_1
    #   value: value1
    # - name: MY_ENV_2
    #   valueFrom:
    #     fieldRef:
    #       fieldPath: status.myEnv2

    ## Custom sidecar containers can be injected into the FE pods,
    ## which can act as a tracing agent or for any other use case
    # additionalContainers:
    # - name: myCustomContainer
    #   image: ubuntu

    ## Custom additional volumes in FE pods.
    ## Ref: https://kubernetes.io/docs/concepts/storage/persistent-volumes/#types-of-persistent-volumes
    # additionalVolumes:
    # - name: nfs
    #   nfs:
    #     server: 192.168.0.2
    #     path: /nfs

    ## Custom additional volume mounts in FE pods.
    # additionalVolumeMounts:
    # # this must match `name` in `additionalVolumes`
    # - name: nfs
    #   mountPath: /nfs

    ## The following block overwrites cluster-level configurations in `spec`
    # serviceAccount: ""
    # affinity: {}
    # tolerations: {}
    # priorityClassName: ""
    # statefulSetUpdateStrategy: RollingUpdate
    # nodeSelector:
    #   app.kubernetes.io/component: fe


  ###############################
  # BE(mixed mode) Configuration #
  ###############################
  # When this "be" configuration key is not set, the Doris BE component will not be deployed,
  # and the BE components on the cluster will be deleted (but the pvc for be persistent data
  # will be retained).

  be:
    #########################
    # BE Basic Configuration #
    #########################

    ## Base image of the BE component
    baseImage: ghcr.io/linsoss/doris-be

    ## The replica of the BE component
    replicas: 3

    ## Extra BE config, see: https://doris.apache.org/docs/dev/admin-manual/config/be-config
    # config:
    #  external_table_connect_timeout_sec: '30 seconds'
    #  mem_limit: '90%'

    ## Describes the resource requirements. For production environments, please refer to: https://doris.apache.org/docs/dev/install/standard-deployment/#production-environment
    ## Ref: https://kubernetes.io/docs/concepts/configuration/manage-compute-resources-container/
    requests:
      cpu: '8'
      memory: 32Gi
      storage: 500Gi
    # limits:
    #   cpu: '32'
    #   memory: 64Gi

    ## The storageClassName of the persistent volume for FE persistent data.
    ## If storageClassName is not set, the default Storage Class of the Kubernetes cluster will be used.
    # storageClassName: local

    ############################
    # BE Advanced Configuration #
    ############################

    ## The custom storage of BE used to support cold and hot storage separation.
    ## Ref: https://doris.apache.org/docs/1.2/install/standard-deployment/#deploy-be
    ##    name: custom storage name
    ##    medium: storage medium, SSD(hot storage) or HDD(cold storage)
    ##    request: storage capacity, e.g. "500Gi"
    ##    storageClassName: k8s storage class name for the pvc
    # storage:
    #  - name: storage-cold-1
    #    medium: HDD
    #    request: 500Gi
    #    storageClassName: hdd-pool
    #  - name: storage-cold-2
    #    medium: HDD
    #    request: 500Gi
    #    storageClassName: hdd-pool
    #  - name: storage-hot
    #    medium: SSD
    #    request: 200Gi
    #    storageClassName: ssd-pool

    ## Whether to retain the default data storage mount for BE which is located at be/storage,
    # retainDefaultStorage: false

    ## Annotations for BE pods
    # annotations: {}

    ## Host aliases for BE pods, it will be merged with the hadoopConf field
    ## Ref: https://kubernetes.io/docs/concepts/services-networking/add-entries-to-pod-etc-hosts-with-host-aliases/
    # hostAliases:
    #  - ip: 10.233.123.122
    #    hostnames:
    #      - bg01
    #  - ip: 10.233.123.123
    #    hostnames:
    #      - bg02

    ## List of environment variables to set in the container
    ## Ref: https://kubernetes.io/docs/tasks/inject-data-application/environment-variable-expose-pod-information/
    # additionalEnvs:
    # - name: MY_ENV_1
    #   value: value1
    # - name: MY_ENV_2
    #   valueFrom:
    #     fieldRef:
    #       fieldPath: status.myEnv2

    ## Custom sidecar containers can be injected into the BE pods,
    ## which can act as a tracing agent or for any other use case
    # additionalContainers:
    # - name: myCustomContainer
    #   image: ubuntu

    ## Custom additional volumes in BE pods.
    ## Ref: https://kubernetes.io/docs/concepts/storage/persistent-volumes/#types-of-persistent-volumes
    # additionalVolumes:
    # - name: nfs
    #   nfs:
    #     server: 192.168.0.2
    #     path: /nfs

    ## Custom additional volume mounts in BE pods.
    # additionalVolumeMounts:
    # # this must match `name` in `additionalVolumes`
    # - name: nfs
    #   mountPath: /nfs

    ## The following block overwrites cluster-level configurations in `spec`
    # serviceAccount: ""
    # affinity: {}
    # tolerations: {}
    # priorityClassName: ""
    # statefulSetUpdateStrategy: RollingUpdate
    # nodeSelector:
    #   app.kubernetes.io/component: be


  #####################################
  # BE(computation mode) Configuration #
  #####################################
  # When this "cn" configuration key is not set, the Doris CN component will not be deployed,
  # and the CN components on the cluster will be deleted.

  cn:
    #########################
    # CN Basic Configuration #
    #########################

    ## Base image of the CN component
    baseImage: ghcr.io/linsoss/doris-cn

    ## The replica of the CN component
    ## When there is a DorisAutoscaler bound to the current DorisCluster, cn.replicas will not take effect.
    ## The actual number of replicas adjusted by DorisAutoscaler shall prevail.
    replicas: 2

    ## Extra BE config, see: https://doris.apache.org/docs/dev/admin-manual/config/be-config
    # config:
    #  external_table_connect_timeout_sec: '30 seconds'
    #  mem_limit: '90%'

    ## The resource requirements. For production environments, please refer to: https://doris.apache.org/docs/dev/install/standard-deployment/#production-environment
    ## Ref: https://kubernetes.io/docs/concepts/configuration/manage-compute-resources-container/
    requests:
      cpu: '8'
      memory: 32Gi
    # limits:
    #   cpu: '32'
    #   memory: 64Gi

    ############################
    # CN Advanced Configuration #
    ############################

    ## Annotations for CN pods
    # annotations: {}

    ## Host aliases for BE pods, it will be merged with the hadoopConf field
    ## Ref: https://kubernetes.io/docs/concepts/services-networking/add-entries-to-pod-etc-hosts-with-host-aliases/
    # hostAliases:
    #  - ip: 10.233.123.122
    #    hostnames:
    #      - bg01
    #  - ip: 10.233.123.123
    #    hostnames:
    #      - bg02

    ## List of environment variables to set in the container
    ## Ref: https://kubernetes.io/docs/tasks/inject-data-application/environment-variable-expose-pod-information/
    # additionalEnvs:
    # - name: MY_ENV_1
    #   value: value1
    # - name: MY_ENV_2
    #   valueFrom:
    #     fieldRef:
    #       fieldPath: status.myEnv2

    ## Custom sidecar containers can be injected into the BE pods,
    ## which can act as a tracing agent or for any other use case
    # additionalContainers:
    # - name: myCustomContainer
    #   image: ubuntu

    ## Custom additional volumes in BE pods.
    ## Ref: https://kubernetes.io/docs/concepts/storage/persistent-volumes/#types-of-persistent-volumes
    # additionalVolumes:
    # - name: nfs
    #   nfs:
    #     server: 192.168.0.2
    #     path: /nfs

    ## Custom additional volume mounts in BE pods.
    # additionalVolumeMounts:
    # # this must match `name` in `additionalVolumes`
    # - name: nfs
    #   mountPath: /nfs

    ## The following block overwrites cluster-level configurations in `spec`
    # serviceAccount: ""
    # affinity: {}
    # tolerations: {}
    # priorityClassName: ""
    # statefulSetUpdateStrategy: RollingUpdate
    # nodeSelector:
    #   app.kubernetes.io/component: cn


  #######################
  # Broker Configuration #
  #######################
  # When this "broker" configuration key is not set, the Doris broker component will not be deployed,
  # and the broker components on the cluster will be deleted.

  broker:
    #############################
    # Broker Basic Configuration #
    #############################

    ## Base image of the Broker component
    baseImage: ghcr.io/linsoss/doris-broker

    ## The replica of the Broker component
    replicas: 1

    ## Extra Broker config.
    config: { }

    ## Describe the resource requirements. For production environments, please refer to: https://doris.apache.org/docs/dev/install/standard-deployment/#production-environment.
    ## Ref: https://kubernetes.io/docs/concepts/configuration/manage-compute-resources-container/
    requests:
      cpu: '1'
      memory: 2Gi
    #    limits:
    #      cpu: 8
    #      memory: 16Gi

    ################################
    # Broker Advanced Configuration #
    ################################

    ## Annotations for Broker pods
    # annotations: {}

    ## Host aliases for BE pods, it will be merged with the hadoopConf field
    ## Ref: https://kubernetes.io/docs/concepts/services-networking/add-entries-to-pod-etc-hosts-with-host-aliases/
    # hostAliases:
    #  - ip: 10.233.123.122
    #    hostnames:
    #      - bg01
    #  - ip: 10.233.123.123
    #    hostnames:
    #      - bg02

    ## List of environment variables to set in the container
    ## Ref: https://kubernetes.io/docs/tasks/inject-data-application/environment-variable-expose-pod-information/
    # additionalEnvs:
    # - name: MY_ENV_1
    #   value: value1
    # - name: MY_ENV_2
    #   valueFrom:
    #     fieldRef:
    #       fieldPath: status.myEnv2

    ## Custom sidecar containers can be injected into the BE pods,
    ## which can act as a tracing agent or for any other use case
    # additionalContainers:
    # - name: myCustomContainer
    #   image: ubuntu

    ## Custom additional volumes in BE pods.
    ## Ref: https://kubernetes.io/docs/concepts/storage/persistent-volumes/#types-of-persistent-volumes
    # additionalVolumes:
    # - name: nfs
    #   nfs:
    #     server: 192.168.0.2
    #     path: /nfs

    ## Custom additional volume mounts in BE pods.
    # additionalVolumeMounts:
    # # this must match `name` in `additionalVolumes`
    # - name: nfs
    #   mountPath: /nfs

    ## The following block overwrites cluster-level configurations in `spec`
    # serviceAccount: ""
    # affinity: {}
    # tolerations: {}
    # priorityClassName: ""
    # statefulSetUpdateStrategy: RollingUpdate
    # nodeSelector:
    #   app.kubernetes.io/component: broker

Note

建议在 ${cluster_name} 目录下组织 Doris 集群的配置，并将其另存为 ${cluster_name}/doris-cluster.yaml。修改配置并提交后，会自动应用到 Doris 集群中。

集群名称

通过更改 DorisCuster CR 中的 metadata.name 来配置集群名称。

版本

正常情况下，集群内的各组件应该使用相同版本，所以一般建议配置 spec.<fe/be/cn/broker>.baseImage + spec.version 即可。

Component	Image
FE	ghcr.io/linsoss/doris-fe
BE	ghcr.io/linsoss/doris-be
CN	ghcr.io/linsoss/doris-cn
Broker	ghcr.io/linsoss/doris-broker

存储

如果需要设置存储类型，可以修改 ${cluster_name}/doris-cluster.yaml 中各组件的 storageClassName 字段。

Doris 集群不同组件对磁盘的要求不一样，所以部署集群前，要根据当前 Kubernetes 集群支持的存储类型以及使用场景，参考存储配置文档为 Doris 集群各组件选择合适的存储类型。

如果需要为 Doris BE 配置冷热存储分离存储，可以参考配置 Doris BE 冷热分离存储。

Doris 组件配置参数

可以通过 spec.<fe/be/cn/broker>.config 来配置各个组件的参数。

比如想修改 FE 以下配置参数：

prefer_compute_node_for_external_table=true
enable_spark_load=true

则修改 DorisCluster 的以下配置：

spec:
  fe:
    config:
      prefer_compute_node_for_external_table: 'true'
      enable_spark_load: 'true'

Note

并不需要为 FE 设置 enable_fqdn_mode，Doris Operator 会强制自动将该参数设置为 true 并注入容器。

配置 Doris 服务

通过配置 spec.fe.service 定义不同的 Service 类型，如 ClusterIP 、 NodePort。默认情况下 Doris Operator 会为 FE 创建一个额外的 ClusterIP 类型 Service。

ClusterIP
ClusterIP 是通过集群的内部 IP 暴露服务，选择该类型的服务时，只能在集群内部访问，使用 ClusterIP 或者 Service 域名（${cluster_name}-fe.${namespace}）访问。
```
spec:
  doris:
    service:
      type: ClusterIP
```
NodePort
在本地测试时候，可选择通过 NodePort 暴露，Doris Operator 会绑定 FE 的 SQL 查询端口和 Web UI 端口到 NodePort。
NodePort 是通过节点的 IP 和静态端口暴露服务。通过请求 NodeIP + NodePort，可以从集群的外部访问一个 NodePort 服务。
```
spec:
  doris:
    service:
      type: NodePort
```

Hadoop 连接配置

当 Doris 集群需要连接 Hadoop，相关的 Hadoop 配置文件是必不可少的，spec.hadoopConf 配置项提供了方便的向 FE、BE、CN、Broke 注入 Hadoop 配置的方式。

spec:
  hadoopConf:
  # Hadoop 集群的 hostname-ip 映射
  hosts:
    - ip: 10.233.123.189
      name: hadoop-01
    - ip: 10.233.123.179
      name: hadoop-02
    - ip: 10.233.123.179
      name: hadoop-03
  # Hadoop 配置文件内容
  configs:
    hdfs-site.xml: |
      <configuration>
      ...
      </configuration>      
    hive-site.xml: |
      <configuration>
      ...
      </configuration>

物理拓扑高可用

Doris 是一个分布式数据库，以下介绍 3 种方式来为维持 Doris 在 Kubernetes 上的物理拓扑高可用。

通过 nodeSelector 约束调度实例

通过各组件配置的 nodeSelector 字段，可以约束组件的实例只能调度到特定的节点上。关于 nodeSelector 的更多说明，请参阅 nodeSelector。

apiVersion: al-assad.github.io/v1beta1
kind: DorisCluster
# ...
spec:
  fe:
    nodeSelector:
      node-role.kubernetes.io/fe: true
    # ...
  be:
    nodeSelector:
      node-role.kubernetes.io/be: true
    # ...
  cn:
    nodeSelector:
      node-role.kubernetes.io/cn: true
    # ...
  broker:
    nodeSelector:
      node-role.kubernetes.io/broker: true

通过 tolerations 调度实例

通过各组件配置的 tolerations 字段，可以允许组件的实例能够调度到带有与之匹配的污点 ( Taint) 的节点上。关于污点与容忍度的更多说明，请参阅 Taints and Tolerations。

apiVersion: al-assad.github.io/v1beta1
kind: DorisCluster
# ...
spec:
  fe:
    tolerations:
      - effect: NoSchedule
        key: dedicated
        operator: Equal
        value: fe
    # ...
  be:
    tolerations:
      - effect: NoSchedule
        key: dedicated
        operator: Equal
        value: be
    # ...
  cn:
    tolerations:
      - effect: NoSchedule
        key: dedicated
        operator: Equal
        value: cn
    # ...
  broker:
    tolerations:
      - effect: NoSchedule
        key: dedicated
        operator: Equal
        value: broker
    # ...

通过 affinity 调度实例

配置 PodAntiAffinity 能尽量避免同一组件的不同实例部署到同一个物理拓扑节点上，从而达到高可用的目的。关于 Affinity 的使用说明，请参阅 Affinity & AntiAffinity。

下面是一个避免 FE 实例调度到同一个物理节点的例子：

affinity:
  podAntiAffinity:
    requiredDuringSchedulingIgnoredDuringExecution:
      - labelSelector:
          matchExpressions:
            - key: app.kubernetes.io/component
              operator: In
              values:
                - fe
            - key: app.kubernetes.io/instance
              operator: In
              values:
                - ${name}
        topologyKey: kubernetes.io/hostname

Edit this page on GitHub

配置 Storage Class

部署 Doris 集群

Docs

Doris Operator

Title here

配置 Doris 集群

资源配置

部署配置

集群名称

版本

存储

Doris 组件配置参数

配置 Doris 服务

Hadoop 连接配置

物理拓扑高可用

通过 nodeSelector 约束调度实例

通过 tolerations 调度实例

通过 affinity 调度实例

配置 Doris 集群

资源配置#

部署配置#

集群名称#

版本#

存储#

Doris 组件配置参数#

配置 Doris 服务#

Hadoop 连接配置#

物理拓扑高可用#

通过 nodeSelector 约束调度实例#

通过 tolerations 调度实例#

通过 affinity 调度实例#

资源配置

部署配置

集群名称

版本

存储

Doris 组件配置参数

配置 Doris 服务

Hadoop 连接配置

物理拓扑高可用

通过 nodeSelector 约束调度实例

通过 tolerations 调度实例

通过 affinity 调度实例