快速试用 Kubernetes 部署

Kubernetes部署目的是在Kubernetes集群中部署 DolphinScheduler 服务,能调度大量任务,可用于在生产中部署。

如果你是新手,想要体验 DolphinScheduler 的功能,推荐使用Standalone方式体检。如果你想体验更完整的功能,或者更大的任务量,推荐使用伪集群部署。如果你是在生产中使用,推荐使用集群部署或者kubernetes

先决条件

安装 dolphinscheduler

请下载源码包 apache-dolphinscheduler-2.0.0-src.tar.gz,下载地址: 下载

发布一个名为 dolphinscheduler 的版本(release),请执行以下命令:

  1. $ tar -zxvf apache-dolphinscheduler-2.0.0-src.tar.gz
  2. $ cd apache-dolphinscheduler-2.0.0-src/docker/kubernetes/dolphinscheduler
  3. $ helm repo add bitnami https://charts.bitnami.com/bitnami
  4. $ helm dependency update .
  5. $ helm install dolphinscheduler . --set image.tag=2.0.0

将名为 dolphinscheduler 的版本(release) 发布到 test 的命名空间中:

  1. $ helm install dolphinscheduler . -n test

提示: 如果名为 test 的命名空间被使用, 选项参数 -n test 需要添加到 helmkubectl 命令中

这些命令以默认配置在 Kubernetes 集群上部署 DolphinScheduler,附录-配置部分列出了可以在安装过程中配置的参数

提示: 列出所有已发布的版本,使用 helm list

PostgreSQL (用户 root, 密码 root, 数据库 dolphinscheduler) 和 ZooKeeper 服务将会默认启动

访问 DolphinScheduler 前端页面

如果 values.yaml 文件中的 ingress.enabled 被设置为 true, 在浏览器中访问 http://${ingress.host}/dolphinscheduler 即可

提示: 如果 ingress 访问遇到问题,请联系 Kubernetes 管理员并查看 Ingress

否则,当 api.service.type=ClusterIP 时,你需要执行 port-forward 端口转发命令:

  1. $ kubectl port-forward --address 0.0.0.0 svc/dolphinscheduler-api 12345:12345
  2. $ kubectl port-forward --address 0.0.0.0 -n test svc/dolphinscheduler-api 12345:12345 # 使用 test 命名空间

提示: 如果出现 unable to do port forwarding: socat not found 错误, 需要先安装 socat

然后访问前端页面: http://192.168.xx.xx:12345/dolphinscheduler (本地地址为 http://127.0.0.1:12345/dolphinscheduler)

或者当 api.service.type=NodePort 时,你需要执行命令:

  1. NODE_IP=$(kubectl get no -n {{ .Release.Namespace }} -o jsonpath="{.items[0].status.addresses[0].address}")
  2. NODE_PORT=$(kubectl get svc {{ template "dolphinscheduler.fullname" . }}-api -n {{ .Release.Namespace }} -o jsonpath="{.spec.ports[0].nodePort}")
  3. echo http://$NODE_IP:$NODE_PORT/dolphinscheduler

然后访问前端页面: http://192.168.xx.xx:12345/dolphinscheduler

默认的用户是admin,默认的密码是dolphinscheduler123

请参考用户手册章节的快速上手查看如何使用DolphinScheduler

卸载 dolphinscheduler

卸载名为 dolphinscheduler 的版本(release),请执行:

  1. $ helm uninstall dolphinscheduler

该命令将删除与 dolphinscheduler 相关的所有 Kubernetes 组件(但PVC除外),并删除版本(release)

要删除与 dolphinscheduler 相关的PVC,请执行:

  1. $ kubectl delete pvc -l app.kubernetes.io/instance=dolphinscheduler

注意: 删除PVC也会删除所有数据,请谨慎操作!

配置

配置文件为 values.yaml附录-配置 表格列出了 DolphinScheduler 的可配置参数及其默认值

支持矩阵

Type支持备注
Shell
Python2
Python3间接支持详见 FAQ
Hadoop2间接支持详见 FAQ
Hadoop3尚未确定尚未测试
Spark-Local(client)间接支持详见 FAQ
Spark-YARN(cluster)间接支持详见 FAQ
Spark-Standalone(cluster)尚不
Spark-Kubernetes(cluster)尚不
Flink-Local(local>=1.11)尚不Generic CLI 模式尚未支持
Flink-YARN(yarn-cluster)间接支持详见 FAQ
Flink-YARN(yarn-session/yarn-per-job/yarn-application>=1.11)尚不Generic CLI 模式尚未支持
Flink-Standalone(default)尚不
Flink-Standalone(remote>=1.11)尚不Generic CLI 模式尚未支持
Flink-Kubernetes(default)尚不
Flink-Kubernetes(remote>=1.11)尚不Generic CLI 模式尚未支持
Flink-NativeKubernetes(kubernetes-session/application>=1.11)尚不Generic CLI 模式尚未支持
MapReduce间接支持详见 FAQ
Kerberos间接支持详见 FAQ
HTTP
DataX间接支持详见 FAQ
Sqoop间接支持详见 FAQ
SQL-MySQL间接支持详见 FAQ
SQL-PostgreSQL
SQL-Hive间接支持详见 FAQ
SQL-Spark间接支持详见 FAQ
SQL-ClickHouse间接支持详见 FAQ
SQL-Oracle间接支持详见 FAQ
SQL-SQLServer间接支持详见 FAQ
SQL-DB2间接支持详见 FAQ

FAQ

如何查看一个 pod 容器的日志?

列出所有 pods (别名 po):

  1. kubectl get po
  2. kubectl get po -n test # with test namespace

查看名为 dolphinscheduler-master-0 的 pod 容器的日志:

  1. kubectl logs dolphinscheduler-master-0
  2. kubectl logs -f dolphinscheduler-master-0 # 跟随日志输出
  3. kubectl logs --tail 10 dolphinscheduler-master-0 -n test # 显示倒数10行日志

如何在 Kubernetes 上扩缩容 api, master 和 worker?

列出所有 deployments (别名 deploy):

  1. kubectl get deploy
  2. kubectl get deploy -n test # with test namespace

扩缩容 api 至 3 个副本:

  1. kubectl scale --replicas=3 deploy dolphinscheduler-api
  2. kubectl scale --replicas=3 deploy dolphinscheduler-api -n test # with test namespace

列出所有 statefulsets (别名 sts):

  1. kubectl get sts
  2. kubectl get sts -n test # with test namespace

扩缩容 master 至 2 个副本:

  1. kubectl scale --replicas=2 sts dolphinscheduler-master
  2. kubectl scale --replicas=2 sts dolphinscheduler-master -n test # with test namespace

扩缩容 worker 至 6 个副本:

  1. kubectl scale --replicas=6 sts dolphinscheduler-worker
  2. kubectl scale --replicas=6 sts dolphinscheduler-worker -n test # with test namespace

如何用 MySQL 替代 PostgreSQL 作为 DolphinScheduler 的数据库?

由于商业许可证的原因,我们不能直接使用 MySQL 的驱动包.

如果你要使用 MySQL, 你可以基于官方镜像 apache/dolphinscheduler 进行构建.

  1. 下载 MySQL 驱动包 mysql-connector-java-8.0.16.jar
  2. 创建一个新的 Dockerfile,用于添加 MySQL 的驱动包:
  1. FROM dolphinscheduler.docker.scarf.sh/apache/dolphinscheduler:2.0.0
  2. COPY mysql-connector-java-8.0.16.jar /opt/dolphinscheduler/lib
  1. 构建一个包含 MySQL 驱动包的新镜像:
  1. docker build -t apache/dolphinscheduler:mysql-driver .
  1. 推送 docker 镜像 apache/dolphinscheduler:mysql-driver 到一个 docker registry 中

  2. 修改 values.yaml 文件中 image 的 repository 字段,并更新 tagmysql-driver

  3. 修改 values.yaml 文件中 postgresql 的 enabledfalse

  4. 修改 values.yaml 文件中的 externalDatabase 配置 (尤其修改 host, usernamepassword)

  1. externalDatabase:
  2. type: "mysql"
  3. driver: "com.mysql.jdbc.Driver"
  4. host: "localhost"
  5. port: "3306"
  6. username: "root"
  7. password: "root"
  8. database: "dolphinscheduler"
  9. params: "useUnicode=true&characterEncoding=UTF-8"
  1. 部署 dolphinscheduler (详见安装 dolphinscheduler)

如何在数据源中心支持 MySQL 数据源?

由于商业许可证的原因,我们不能直接使用 MySQL 的驱动包.

如果你要添加 MySQL 数据源, 你可以基于官方镜像 apache/dolphinscheduler 进行构建.

  1. 下载 MySQL 驱动包 mysql-connector-java-8.0.16.jar (要求 >=8.0.1)

  2. 创建一个新的 Dockerfile,用于添加 MySQL 驱动包:

  1. FROM dolphinscheduler.docker.scarf.sh/apache/dolphinscheduler:2.0.0
  2. COPY mysql-connector-java-8.0.16.jar /opt/dolphinscheduler/lib
  1. 构建一个包含 MySQL 驱动包的新镜像:
  1. docker build -t apache/dolphinscheduler:mysql-driver .
  1. 推送 docker 镜像 apache/dolphinscheduler:mysql-driver 到一个 docker registry 中

  2. 修改 values.yaml 文件中 image 的 repository 字段,并更新 tagmysql-driver

  3. 部署 dolphinscheduler (详见安装 dolphinscheduler)

  4. 在数据源中心添加一个 MySQL 数据源

如何在数据源中心支持 Oracle 数据源?

由于商业许可证的原因,我们不能直接使用 Oracle 的驱动包.

如果你要添加 Oracle 数据源, 你可以基于官方镜像 apache/dolphinscheduler 进行构建.

  1. 下载 Oracle 驱动包 ojdbc8.jar (例如 ojdbc8-19.9.0.0.jar)

  2. 创建一个新的 Dockerfile,用于添加 Oracle 驱动包:

  1. FROM dolphinscheduler.docker.scarf.sh/apache/dolphinscheduler:2.0.0
  2. COPY ojdbc8-19.9.0.0.jar /opt/dolphinscheduler/lib
  1. 构建一个包含 Oracle 驱动包的新镜像:
  1. docker build -t apache/dolphinscheduler:oracle-driver .
  1. 推送 docker 镜像 apache/dolphinscheduler:oracle-driver 到一个 docker registry 中

  2. 修改 values.yaml 文件中 image 的 repository 字段,并更新 tagoracle-driver

  3. 部署 dolphinscheduler (详见安装 dolphinscheduler)

  4. 在数据源中心添加一个 Oracle 数据源

如何支持 Python 2 pip 以及自定义 requirements.txt?

  1. 创建一个新的 Dockerfile,用于安装 pip:
  1. FROM dolphinscheduler.docker.scarf.sh/apache/dolphinscheduler:2.0.0
  2. COPY requirements.txt /tmp
  3. RUN apt-get update && \
  4. apt-get install -y --no-install-recommends python-pip && \
  5. pip install --no-cache-dir -r /tmp/requirements.txt && \
  6. rm -rf /var/lib/apt/lists/*

这个命令会安装默认的 pip 18.1. 如果你想升级 pip, 只需添加一行

  1. pip install --no-cache-dir -U pip && \
  1. 构建一个包含 pip 的新镜像:
  1. docker build -t apache/dolphinscheduler:pip .
  1. 推送 docker 镜像 apache/dolphinscheduler:pip 到一个 docker registry 中

  2. 修改 values.yaml 文件中 image 的 repository 字段,并更新 tagpip

  3. 部署 dolphinscheduler (详见安装 dolphinscheduler)

  4. 在一个新 Python 任务下验证 pip

如何支持 Python 3?

  1. 创建一个新的 Dockerfile,用于安装 Python 3:
  1. FROM dolphinscheduler.docker.scarf.sh/apache/dolphinscheduler:2.0.0
  2. RUN apt-get update && \
  3. apt-get install -y --no-install-recommends python3 && \
  4. rm -rf /var/lib/apt/lists/*

这个命令会安装默认的 Python 3.7.3. 如果你也想安装 pip3, 将 python3 替换为 python3-pip 即可

  1. apt-get install -y --no-install-recommends python3-pip && \
  1. 构建一个包含 Python 3 的新镜像:
  1. docker build -t apache/dolphinscheduler:python3 .
  1. 推送 docker 镜像 apache/dolphinscheduler:python3 到一个 docker registry 中

  2. 修改 values.yaml 文件中 image 的 repository 字段,并更新 tagpython3

  3. 修改 values.yaml 文件中的 PYTHON_HOME/usr/bin/python3

  4. 部署 dolphinscheduler (详见安装 dolphinscheduler)

  5. 在一个新 Python 任务下验证 Python 3

以 Spark 2.4.7 为例:

  1. 下载 Spark 2.4.7 发布的二进制包 spark-2.4.7-bin-hadoop2.7.tgz

  2. 确保 common.sharedStoragePersistence.enabled 开启

  3. 部署 dolphinscheduler (详见安装 dolphinscheduler)

  4. 复制 Spark 3.1.1 二进制包到 Docker 容器中

  1. kubectl cp spark-2.4.7-bin-hadoop2.7.tgz dolphinscheduler-worker-0:/opt/soft
  2. kubectl cp -n test spark-2.4.7-bin-hadoop2.7.tgz dolphinscheduler-worker-0:/opt/soft # with test namespace

因为存储卷 sharedStoragePersistence 被挂载到 /opt/soft, 因此 /opt/soft 中的所有文件都不会丢失

  1. 登录到容器并确保 SPARK_HOME2 存在
  1. kubectl exec -it dolphinscheduler-worker-0 bash
  2. kubectl exec -n test -it dolphinscheduler-worker-0 bash # with test namespace
  3. cd /opt/soft
  4. tar zxf spark-2.4.7-bin-hadoop2.7.tgz
  5. rm -f spark-2.4.7-bin-hadoop2.7.tgz
  6. ln -s spark-2.4.7-bin-hadoop2.7 spark2 # or just mv
  7. $SPARK_HOME2/bin/spark-submit --version

如果一切执行正常,最后一条命令将会打印 Spark 版本信息

  1. 在一个 Shell 任务下验证 Spark
  1. $SPARK_HOME2/bin/spark-submit --class org.apache.spark.examples.SparkPi $SPARK_HOME2/examples/jars/spark-examples_2.11-2.4.7.jar

检查任务日志是否包含输出 Pi is roughly 3.146015

  1. 在一个 Spark 任务下验证 Spark

文件 spark-examples_2.11-2.4.7.jar 需要先被上传到资源中心,然后创建一个 Spark 任务并设置:

  • Spark版本: SPARK2
  • 主函数的Class: org.apache.spark.examples.SparkPi
  • 主程序包: spark-examples_2.11-2.4.7.jar
  • 部署方式: local

同样地, 检查任务日志是否包含输出 Pi is roughly 3.146015

  1. 验证 Spark on YARN

Spark on YARN (部署方式为 clusterclient) 需要 Hadoop 支持. 类似于 Spark 支持, 支持 Hadoop 的操作几乎和前面的步骤相同

确保 $HADOOP_HOME$HADOOP_CONF_DIR 存在

如何支持 Spark 3?

事实上,使用 spark-submit 提交应用的方式是相同的, 无论是 Spark 1, 2 或 3. 换句话说,SPARK_HOME2 的语义是第二个 SPARK_HOME, 而非 SPARK2HOME, 因此只需设置 SPARK_HOME2=/path/to/spark3 即可

以 Spark 3.1.1 为例:

  1. 下载 Spark 3.1.1 发布的二进制包 spark-3.1.1-bin-hadoop2.7.tgz

  2. 确保 common.sharedStoragePersistence.enabled 开启

  3. 部署 dolphinscheduler (详见安装 dolphinscheduler)

  4. 复制 Spark 3.1.1 二进制包到 Docker 容器中

  1. kubectl cp spark-3.1.1-bin-hadoop2.7.tgz dolphinscheduler-worker-0:/opt/soft
  2. kubectl cp -n test spark-3.1.1-bin-hadoop2.7.tgz dolphinscheduler-worker-0:/opt/soft # with test namespace
  1. 登录到容器并确保 SPARK_HOME2 存在
  1. kubectl exec -it dolphinscheduler-worker-0 bash
  2. kubectl exec -n test -it dolphinscheduler-worker-0 bash # with test namespace
  3. cd /opt/soft
  4. tar zxf spark-3.1.1-bin-hadoop2.7.tgz
  5. rm -f spark-3.1.1-bin-hadoop2.7.tgz
  6. ln -s spark-3.1.1-bin-hadoop2.7 spark2 # or just mv
  7. $SPARK_HOME2/bin/spark-submit --version

如果一切执行正常,最后一条命令将会打印 Spark 版本信息

  1. 在一个 Shell 任务下验证 Spark
  1. $SPARK_HOME2/bin/spark-submit --class org.apache.spark.examples.SparkPi $SPARK_HOME2/examples/jars/spark-examples_2.12-3.1.1.jar

检查任务日志是否包含输出 Pi is roughly 3.146015

如何在 Master、Worker 和 Api 服务之间支持共享存储?

例如, Master、Worker 和 Api 服务可能同时使用 Hadoop

  1. 修改 values.yaml 文件中下面的配置项
  1. common:
  2. sharedStoragePersistence:
  3. enabled: false
  4. mountPath: "/opt/soft"
  5. accessModes:
  6. - "ReadWriteMany"
  7. storageClassName: "-"
  8. storage: "20Gi"

storageClassNamestorage 需要被修改为实际值

注意: storageClassName 必须支持访问模式: ReadWriteMany

  1. 将 Hadoop 复制到目录 /opt/soft

  2. 确保 $HADOOP_HOME$HADOOP_CONF_DIR 正确

如何支持本地文件存储而非 HDFS 和 S3?

修改 values.yaml 文件中下面的配置项

  1. common:
  2. configmap:
  3. RESOURCE_STORAGE_TYPE: "HDFS"
  4. RESOURCE_UPLOAD_PATH: "/dolphinscheduler"
  5. FS_DEFAULT_FS: "file:///"
  6. fsFileResourcePersistence:
  7. enabled: true
  8. accessModes:
  9. - "ReadWriteMany"
  10. storageClassName: "-"
  11. storage: "20Gi"

storageClassNamestorage 需要被修改为实际值

注意: storageClassName 必须支持访问模式: ReadWriteMany

如何支持 S3 资源存储,例如 MinIO?

以 MinIO 为例: 修改 values.yaml 文件中下面的配置项

  1. common:
  2. configmap:
  3. RESOURCE_STORAGE_TYPE: "S3"
  4. RESOURCE_UPLOAD_PATH: "/dolphinscheduler"
  5. FS_DEFAULT_FS: "s3a://BUCKET_NAME"
  6. FS_S3A_ENDPOINT: "http://MINIO_IP:9000"
  7. FS_S3A_ACCESS_KEY: "MINIO_ACCESS_KEY"
  8. FS_S3A_SECRET_KEY: "MINIO_SECRET_KEY"

BUCKET_NAME, MINIO_IP, MINIO_ACCESS_KEYMINIO_SECRET_KEY 需要被修改为实际值

注意: MINIO_IP 只能使用 IP 而非域名, 因为 DolphinScheduler 尚不支持 S3 路径风格访问 (S3 path style access)

如何配置 SkyWalking?

修改 values.yaml 文件中的 SKYWALKING 配置项

  1. common:
  2. configmap:
  3. SKYWALKING_ENABLE: "true"
  4. SW_AGENT_COLLECTOR_BACKEND_SERVICES: "127.0.0.1:11800"
  5. SW_GRPC_LOG_SERVER_HOST: "127.0.0.1"
  6. SW_GRPC_LOG_SERVER_PORT: "11800"

附录-配置

ParameterDescriptionDefault
timezoneWorld time and date for cities in all time zonesAsia/Shanghai
image.repositoryDocker image repository for the DolphinSchedulerapache/dolphinscheduler
image.tagDocker image version for the DolphinSchedulerlatest
image.pullPolicyImage pull policy. One of Always, Never, IfNotPresentIfNotPresent
image.pullSecretImage pull secret. An optional reference to secret in the same namespace to use for pulling any of the imagesnil
postgresql.enabledIf not exists external PostgreSQL, by default, the DolphinScheduler will use a internal PostgreSQLtrue
postgresql.postgresqlUsernameThe username for internal PostgreSQLroot
postgresql.postgresqlPasswordThe password for internal PostgreSQLroot
postgresql.postgresqlDatabaseThe database for internal PostgreSQLdolphinscheduler
postgresql.persistence.enabledSet postgresql.persistence.enabled to true to mount a new volume for internal PostgreSQLfalse
postgresql.persistence.sizePersistentVolumeClaim size20Gi
postgresql.persistence.storageClassPostgreSQL data persistent volume storage class. If set to “-“, storageClassName: “”, which disables dynamic provisioning-
externalDatabase.typeIf exists external PostgreSQL, and set postgresql.enabled value to false. DolphinScheduler’s database type will use itpostgresql
externalDatabase.driverIf exists external PostgreSQL, and set postgresql.enabled value to false. DolphinScheduler’s database driver will use itorg.postgresql.Driver
externalDatabase.hostIf exists external PostgreSQL, and set postgresql.enabled value to false. DolphinScheduler’s database host will use itlocalhost
externalDatabase.portIf exists external PostgreSQL, and set postgresql.enabled value to false. DolphinScheduler’s database port will use it5432
externalDatabase.usernameIf exists external PostgreSQL, and set postgresql.enabled value to false. DolphinScheduler’s database username will use itroot
externalDatabase.passwordIf exists external PostgreSQL, and set postgresql.enabled value to false. DolphinScheduler’s database password will use itroot
externalDatabase.databaseIf exists external PostgreSQL, and set postgresql.enabled value to false. DolphinScheduler’s database database will use itdolphinscheduler
externalDatabase.paramsIf exists external PostgreSQL, and set postgresql.enabled value to false. DolphinScheduler’s database params will use itcharacterEncoding=utf8
zookeeper.enabledIf not exists external Zookeeper, by default, the DolphinScheduler will use a internal Zookeepertrue
zookeeper.fourlwCommandsWhitelistA list of comma separated Four Letter Words commands to usesrvr,ruok,wchs,cons
zookeeper.persistence.enabledSet zookeeper.persistence.enabled to true to mount a new volume for internal Zookeeperfalse
zookeeper.persistence.sizePersistentVolumeClaim size20Gi
zookeeper.persistence.storageClassZookeeper data persistent volume storage class. If set to “-“, storageClassName: “”, which disables dynamic provisioning-
zookeeper.zookeeperRootSpecify dolphinscheduler root directory in Zookeeper/dolphinscheduler
externalZookeeper.zookeeperQuorumIf exists external Zookeeper, and set zookeeper.enabled value to false. Specify Zookeeper quorum127.0.0.1:2181
externalZookeeper.zookeeperRootIf exists external Zookeeper, and set zookeeper.enabled value to false. Specify dolphinscheduler root directory in Zookeeper/dolphinscheduler
common.configmap.DOLPHINSCHEDULER_OPTSThe jvm options for dolphinscheduler, suitable for all servers“”
common.configmap.DATA_BASEDIR_PATHUser data directory path, self configuration, please make sure the directory exists and have read write permissions/tmp/dolphinscheduler
common.configmap.RESOURCE_STORAGE_TYPEResource storage type: HDFS, S3, NONEHDFS
common.configmap.RESOURCE_UPLOAD_PATHResource store on HDFS/S3 path, please make sure the directory exists on hdfs and have read write permissions/dolphinscheduler
common.configmap.FS_DEFAULT_FSResource storage file system like file:///, hdfs://mycluster:8020 or s3a://dolphinschedulerfile:///
common.configmap.FS_S3A_ENDPOINTS3 endpoint when common.configmap.RESOURCE_STORAGE_TYPE is set to S3s3.xxx.amazonaws.com
common.configmap.FS_S3A_ACCESS_KEYS3 access key when common.configmap.RESOURCE_STORAGE_TYPE is set to S3xxxxxxx
common.configmap.FS_S3A_SECRET_KEYS3 secret key when common.configmap.RESOURCE_STORAGE_TYPE is set to S3xxxxxxx
common.configmap.HADOOP_SECURITY_AUTHENTICATION_STARTUP_STATEWhether to startup kerberosfalse
common.configmap.JAVA_SECURITY_KRB5_CONF_PATHThe java.security.krb5.conf path/opt/krb5.conf
common.configmap.LOGIN_USER_KEYTAB_USERNAMEThe login user from keytab usernamehdfs@HADOOP.COM
common.configmap.LOGIN_USER_KEYTAB_PATHThe login user from keytab path/opt/hdfs.keytab
common.configmap.KERBEROS_EXPIRE_TIMEThe kerberos expire time, the unit is hour2
common.configmap.HDFS_ROOT_USERThe HDFS root user who must have the permission to create directories under the HDFS root pathhdfs
common.configmap.RESOURCE_MANAGER_HTTPADDRESS_PORTSet resource manager httpaddress port for yarn8088
common.configmap.YARN_RESOURCEMANAGER_HA_RM_IDSIf resourcemanager HA is enabled, please set the HA IPsnil
common.configmap.YARN_APPLICATION_STATUS_ADDRESSIf resourcemanager is single, you only need to replace ds1 to actual resourcemanager hostname, otherwise keep defaulthttp://ds1:%s/ws/v1/cluster/apps/%s
common.configmap.SKYWALKING_ENABLESet whether to enable skywalkingfalse
common.configmap.SW_AGENT_COLLECTOR_BACKEND_SERVICESSet agent collector backend services for skywalking127.0.0.1:11800
common.configmap.SW_GRPC_LOG_SERVER_HOSTSet grpc log server host for skywalking127.0.0.1
common.configmap.SW_GRPC_LOG_SERVER_PORTSet grpc log server port for skywalking11800
common.configmap.HADOOP_HOMESet HADOOP_HOME for DolphinScheduler’s task environment/opt/soft/hadoop
common.configmap.HADOOP_CONF_DIRSet HADOOP_CONF_DIR for DolphinScheduler’s task environment/opt/soft/hadoop/etc/hadoop
common.configmap.SPARK_HOME1Set SPARK_HOME1 for DolphinScheduler’s task environment/opt/soft/spark1
common.configmap.SPARK_HOME2Set SPARK_HOME2 for DolphinScheduler’s task environment/opt/soft/spark2
common.configmap.PYTHON_HOMESet PYTHON_HOME for DolphinScheduler’s task environment/usr/bin/python
common.configmap.JAVA_HOMESet JAVA_HOME for DolphinScheduler’s task environment/usr/local/openjdk-8
common.configmap.HIVE_HOMESet HIVE_HOME for DolphinScheduler’s task environment/opt/soft/hive
common.configmap.FLINK_HOMESet FLINK_HOME for DolphinScheduler’s task environment/opt/soft/flink
common.configmap.DATAX_HOMESet DATAX_HOME for DolphinScheduler’s task environment/opt/soft/datax
common.sharedStoragePersistence.enabledSet common.sharedStoragePersistence.enabled to true to mount a shared storage volume for Hadoop, Spark binary and etcfalse
common.sharedStoragePersistence.mountPathThe mount path for the shared storage volume/opt/soft
common.sharedStoragePersistence.accessModesPersistentVolumeClaim access modes, must be ReadWriteMany[ReadWriteMany]
common.sharedStoragePersistence.storageClassNameShared Storage persistent volume storage class, must support the access mode: ReadWriteMany-
common.sharedStoragePersistence.storagePersistentVolumeClaim size20Gi
common.fsFileResourcePersistence.enabledSet common.fsFileResourcePersistence.enabled to true to mount a new file resource volume for api and workerfalse
common.fsFileResourcePersistence.accessModesPersistentVolumeClaim access modes, must be ReadWriteMany[ReadWriteMany]
common.fsFileResourcePersistence.storageClassNameResource persistent volume storage class, must support the access mode: ReadWriteMany-
common.fsFileResourcePersistence.storagePersistentVolumeClaim size20Gi
master.podManagementPolicyPodManagementPolicy controls how pods are created during initial scale up, when replacing pods on nodes, or when scaling downParallel
master.replicasReplicas is the desired number of replicas of the given Template3
master.annotationsThe annotations for master server{}
master.affinityIf specified, the pod’s scheduling constraints{}
master.nodeSelectorNodeSelector is a selector which must be true for the pod to fit on a node{}
master.tolerationsIf specified, the pod’s tolerations{}
master.resourcesThe resource limit and request config for master server{}
master.configmap.MASTER_SERVER_OPTSThe jvm options for master server-Xms1g -Xmx1g -Xmn512m
master.configmap.MASTER_EXEC_THREADSMaster execute thread number to limit process instances100
master.configmap.MASTER_EXEC_TASK_NUMMaster execute task number in parallel per process instance20
master.configmap.MASTER_DISPATCH_TASK_NUMMaster dispatch task number per batch3
master.configmap.MASTER_HOST_SELECTORMaster host selector to select a suitable worker, optional values include Random, RoundRobin, LowerWeightLowerWeight
master.configmap.MASTER_HEARTBEAT_INTERVALMaster heartbeat interval, the unit is second10
master.configmap.MASTER_TASK_COMMIT_RETRYTIMESMaster commit task retry times5
master.configmap.MASTER_TASK_COMMIT_INTERVALmaster commit task interval, the unit is second1
master.configmap.MASTER_MAX_CPULOAD_AVGMaster max cpuload avg, only higher than the system cpu load average, master server can schedule-1 (the number of cpu cores 2)
master.configmap.MASTER_RESERVED_MEMORYMaster reserved memory, only lower than system available memory, master server can schedule, the unit is G0.3
master.livenessProbe.enabledTurn on and off liveness probetrue
master.livenessProbe.initialDelaySecondsDelay before liveness probe is initiated30
master.livenessProbe.periodSecondsHow often to perform the probe30
master.livenessProbe.timeoutSecondsWhen the probe times out5
master.livenessProbe.failureThresholdMinimum consecutive successes for the probe3
master.livenessProbe.successThresholdMinimum consecutive failures for the probe1
master.readinessProbe.enabledTurn on and off readiness probetrue
master.readinessProbe.initialDelaySecondsDelay before readiness probe is initiated30
master.readinessProbe.periodSecondsHow often to perform the probe30
master.readinessProbe.timeoutSecondsWhen the probe times out5
master.readinessProbe.failureThresholdMinimum consecutive successes for the probe3
master.readinessProbe.successThresholdMinimum consecutive failures for the probe1
master.persistentVolumeClaim.enabledSet master.persistentVolumeClaim.enabled to true to mount a new volume for masterfalse
master.persistentVolumeClaim.accessModesPersistentVolumeClaim access modes[ReadWriteOnce]
master.persistentVolumeClaim.storageClassNameMaster logs data persistent volume storage class. If set to “-“, storageClassName: “”, which disables dynamic provisioning-
master.persistentVolumeClaim.storagePersistentVolumeClaim size20Gi
worker.podManagementPolicyPodManagementPolicy controls how pods are created during initial scale up, when replacing pods on nodes, or when scaling downParallel
worker.replicasReplicas is the desired number of replicas of the given Template3
worker.annotationsThe annotations for worker server{}
worker.affinityIf specified, the pod’s scheduling constraints{}
worker.nodeSelectorNodeSelector is a selector which must be true for the pod to fit on a node{}
worker.tolerationsIf specified, the pod’s tolerations{}
worker.resourcesThe resource limit and request config for worker server{}
worker.configmap.LOGGER_SERVER_OPTSThe jvm options for logger server-Xms512m -Xmx512m -Xmn256m
worker.configmap.WORKER_SERVER_OPTSThe jvm options for worker server-Xms1g -Xmx1g -Xmn512m
worker.configmap.WORKER_EXEC_THREADSWorker execute thread number to limit task instances100
worker.configmap.WORKER_HEARTBEAT_INTERVALWorker heartbeat interval, the unit is second10
worker.configmap.WORKER_MAX_CPULOAD_AVGWorker max cpuload avg, only higher than the system cpu load average, worker server can be dispatched tasks-1 (the number of cpu cores 2)
worker.configmap.WORKER_RESERVED_MEMORYWorker reserved memory, only lower than system available memory, worker server can be dispatched tasks, the unit is G0.3
worker.configmap.WORKER_GROUPSWorker groupsdefault
worker.livenessProbe.enabledTurn on and off liveness probetrue
worker.livenessProbe.initialDelaySecondsDelay before liveness probe is initiated30
worker.livenessProbe.periodSecondsHow often to perform the probe30
worker.livenessProbe.timeoutSecondsWhen the probe times out5
worker.livenessProbe.failureThresholdMinimum consecutive successes for the probe3
worker.livenessProbe.successThresholdMinimum consecutive failures for the probe1
worker.readinessProbe.enabledTurn on and off readiness probetrue
worker.readinessProbe.initialDelaySecondsDelay before readiness probe is initiated30
worker.readinessProbe.periodSecondsHow often to perform the probe30
worker.readinessProbe.timeoutSecondsWhen the probe times out5
worker.readinessProbe.failureThresholdMinimum consecutive successes for the probe3
worker.readinessProbe.successThresholdMinimum consecutive failures for the probe1
worker.persistentVolumeClaim.enabledSet worker.persistentVolumeClaim.enabled to true to enable persistentVolumeClaim for workerfalse
worker.persistentVolumeClaim.dataPersistentVolume.enabledSet worker.persistentVolumeClaim.dataPersistentVolume.enabled to true to mount a data volume for workerfalse
worker.persistentVolumeClaim.dataPersistentVolume.accessModesPersistentVolumeClaim access modes[ReadWriteOnce]
worker.persistentVolumeClaim.dataPersistentVolume.storageClassNameWorker data persistent volume storage class. If set to “-“, storageClassName: “”, which disables dynamic provisioning-
worker.persistentVolumeClaim.dataPersistentVolume.storagePersistentVolumeClaim size20Gi
worker.persistentVolumeClaim.logsPersistentVolume.enabledSet worker.persistentVolumeClaim.logsPersistentVolume.enabled to true to mount a logs volume for workerfalse
worker.persistentVolumeClaim.logsPersistentVolume.accessModesPersistentVolumeClaim access modes[ReadWriteOnce]
worker.persistentVolumeClaim.logsPersistentVolume.storageClassNameWorker logs data persistent volume storage class. If set to “-“, storageClassName: “”, which disables dynamic provisioning-
worker.persistentVolumeClaim.logsPersistentVolume.storagePersistentVolumeClaim size20Gi
alert.replicasReplicas is the desired number of replicas of the given Template1
alert.strategy.typeType of deployment. Can be “Recreate” or “RollingUpdate”RollingUpdate
alert.strategy.rollingUpdate.maxSurgeThe maximum number of pods that can be scheduled above the desired number of pods25%
alert.strategy.rollingUpdate.maxUnavailableThe maximum number of pods that can be unavailable during the update25%
alert.annotationsThe annotations for alert server{}
alert.affinityIf specified, the pod’s scheduling constraints{}
alert.nodeSelectorNodeSelector is a selector which must be true for the pod to fit on a node{}
alert.tolerationsIf specified, the pod’s tolerations{}
alert.resourcesThe resource limit and request config for alert server{}
alert.configmap.ALERT_SERVER_OPTSThe jvm options for alert server-Xms512m -Xmx512m -Xmn256m
alert.configmap.XLS_FILE_PATHXLS file path/tmp/xls
alert.configmap.MAIL_SERVER_HOSTMail SERVER HOSTnil
alert.configmap.MAIL_SERVER_PORTMail SERVER PORTnil
alert.configmap.MAIL_SENDERMail SENDERnil
alert.configmap.MAIL_USERMail USERnil
alert.configmap.MAIL_PASSWDMail PASSWORDnil
alert.configmap.MAIL_SMTP_STARTTLS_ENABLEMail SMTP STARTTLS enablefalse
alert.configmap.MAIL_SMTP_SSL_ENABLEMail SMTP SSL enablefalse
alert.configmap.MAIL_SMTP_SSL_TRUSTMail SMTP SSL TRUSTnil
alert.configmap.ENTERPRISE_WECHAT_ENABLEEnterprise Wechat enablefalse
alert.configmap.ENTERPRISE_WECHAT_CORP_IDEnterprise Wechat corp idnil
alert.configmap.ENTERPRISE_WECHAT_SECRETEnterprise Wechat secretnil
alert.configmap.ENTERPRISE_WECHAT_AGENT_IDEnterprise Wechat agent idnil
alert.configmap.ENTERPRISE_WECHAT_USERSEnterprise Wechat usersnil
alert.livenessProbe.enabledTurn on and off liveness probetrue
alert.livenessProbe.initialDelaySecondsDelay before liveness probe is initiated30
alert.livenessProbe.periodSecondsHow often to perform the probe30
alert.livenessProbe.timeoutSecondsWhen the probe times out5
alert.livenessProbe.failureThresholdMinimum consecutive successes for the probe3
alert.livenessProbe.successThresholdMinimum consecutive failures for the probe1
alert.readinessProbe.enabledTurn on and off readiness probetrue
alert.readinessProbe.initialDelaySecondsDelay before readiness probe is initiated30
alert.readinessProbe.periodSecondsHow often to perform the probe30
alert.readinessProbe.timeoutSecondsWhen the probe times out5
alert.readinessProbe.failureThresholdMinimum consecutive successes for the probe3
alert.readinessProbe.successThresholdMinimum consecutive failures for the probe1
alert.persistentVolumeClaim.enabledSet alert.persistentVolumeClaim.enabled to true to mount a new volume for alertfalse
alert.persistentVolumeClaim.accessModesPersistentVolumeClaim access modes[ReadWriteOnce]
alert.persistentVolumeClaim.storageClassNameAlert logs data persistent volume storage class. If set to “-“, storageClassName: “”, which disables dynamic provisioning-
alert.persistentVolumeClaim.storagePersistentVolumeClaim size20Gi
api.replicasReplicas is the desired number of replicas of the given Template1
api.strategy.typeType of deployment. Can be “Recreate” or “RollingUpdate”RollingUpdate
api.strategy.rollingUpdate.maxSurgeThe maximum number of pods that can be scheduled above the desired number of pods25%
api.strategy.rollingUpdate.maxUnavailableThe maximum number of pods that can be unavailable during the update25%
api.annotationsThe annotations for api server{}
api.affinityIf specified, the pod’s scheduling constraints{}
api.nodeSelectorNodeSelector is a selector which must be true for the pod to fit on a node{}
api.tolerationsIf specified, the pod’s tolerations{}
api.resourcesThe resource limit and request config for api server{}
api.configmap.API_SERVER_OPTSThe jvm options for api server-Xms512m -Xmx512m -Xmn256m
api.livenessProbe.enabledTurn on and off liveness probetrue
api.livenessProbe.initialDelaySecondsDelay before liveness probe is initiated30
api.livenessProbe.periodSecondsHow often to perform the probe30
api.livenessProbe.timeoutSecondsWhen the probe times out5
api.livenessProbe.failureThresholdMinimum consecutive successes for the probe3
api.livenessProbe.successThresholdMinimum consecutive failures for the probe1
api.readinessProbe.enabledTurn on and off readiness probetrue
api.readinessProbe.initialDelaySecondsDelay before readiness probe is initiated30
api.readinessProbe.periodSecondsHow often to perform the probe30
api.readinessProbe.timeoutSecondsWhen the probe times out5
api.readinessProbe.failureThresholdMinimum consecutive successes for the probe3
api.readinessProbe.successThresholdMinimum consecutive failures for the probe1
api.persistentVolumeClaim.enabledSet api.persistentVolumeClaim.enabled to true to mount a new volume for apifalse
api.persistentVolumeClaim.accessModesPersistentVolumeClaim access modes[ReadWriteOnce]
api.persistentVolumeClaim.storageClassNameapi logs data persistent volume storage class. If set to “-“, storageClassName: “”, which disables dynamic provisioning-
api.persistentVolumeClaim.storagePersistentVolumeClaim size20Gi
api.service.typetype determines how the Service is exposed. Valid options are ExternalName, ClusterIP, NodePort, and LoadBalancerClusterIP
api.service.clusterIPclusterIP is the IP address of the service and is usually assigned randomly by the masternil
api.service.nodePortnodePort is the port on each node on which this service is exposed when type=NodePortnil
api.service.externalIPsexternalIPs is a list of IP addresses for which nodes in the cluster will also accept traffic for this service[]
api.service.externalNameexternalName is the external reference that kubedns or equivalent will return as a CNAME record for this servicenil
api.service.loadBalancerIPloadBalancerIP when service.type is LoadBalancer. LoadBalancer will get created with the IP specified in this fieldnil
api.service.annotationsannotations may need to be set when service.type is LoadBalancer{}
ingress.enabledEnable ingressfalse
ingress.hostIngress hostdolphinscheduler.org
ingress.pathIngress path/dolphinscheduler
ingress.tls.enabledEnable ingress tlsfalse
ingress.tls.secretNameIngress tls secret namedolphinscheduler-tls