本文記錄了在CDH5.16.2集群上集成Dolphin Scheduler 1.3.1的詳細(xì)流程，特別注意一下MySQL數(shù)據(jù)庫(kù)的連接串！

1

文檔編寫(xiě)目的

• 詳細(xì)記錄CDH5上Dolphin Scheduler 1.3.1的部署流程
• 分布式部署Dolphin Scheduler

2

部署環(huán)境和依賴(lài)組件

為了適配CDH5上的Hive版本， 需要對(duì)DS進(jìn)行源碼編譯部署，最后會(huì)提供編譯好的CDH5版本供各位老鐵下載

集群環(huán)境

• CDH 5.16.2
  • HDFS和YARN都是單點(diǎn)
• DS的官網(wǎng)
  • https://dolphinscheduler.apache.org/en-us/

DS依賴(lài)組件

• MySQL：用于存儲(chǔ)dolphin scheduler的元數(shù)據(jù)，也可以使用pg，這里主要是因?yàn)镃DH集群使用的是MySQL
• Zookeeper：使用CDH集群的zk

3

Dolphin Scheduler 1.3.1集群規(guī)劃

DS服務(wù)

master.eights.com

dn1.eights.com

dn2.eights.com

api		√
master	√	√
worker/log	√		√
alert		√

4

源碼編譯

前置條件

• maven
• jdk
• nvm

代碼拉取

git clone https://github.com/apache/incubator-dolphinscheduler.git

切換CDH5分支

git checkout 1.3.1-release; git checkout -b ds-1.3.1-cdh5.16.2;

修改pom

修改根root的hadoop版本,hive版本和version信息，將每一個(gè)模塊的version都調(diào)整為1.3.1-cdh5.16.2

<hadoop.version>2.6.0</hadoop.version> <hive.jdbc.version>1.1.0</hive.jdbc.version> <version>1.3.1-cdh5.16.2</version>

去除mysql包的scope

執(zhí)行編譯命令進(jìn)行源碼編譯

mvn -U clean package -Prelease -Dmaven.test.skip=true

編譯完成后，dolphinscheduler-dist目錄會(huì)生成

apache-dolphinscheduler-incubating-1.3.1-cdh5.16.2-dolphinscheduler-bin.tar.gz包，1.3.1版本的前后端是打在一起的，并沒(méi)有兩個(gè)包.

5

組件部署

準(zhǔn)備工作

• 創(chuàng)建部署用戶(hù)及配置SSH免密

在所有部署機(jī)器上創(chuàng)建部署用戶(hù)，ds在進(jìn)行任務(wù)執(zhí)行的時(shí)候會(huì)以sudo -u [linux-user]的方式來(lái)執(zhí)行作業(yè)，這里采用dscheduler作為部署用戶(hù)。

# 添加部署用戶(hù) useradd dscheduler; # 設(shè)置密碼 echo "dscheduler" | passwd --stdin dscheduler # 配置免密 echo 'dscheduler ALL=(ALL) NOPASSWD: NOPASSWD: ALL' >> /etc/sudoers# 切換到部署用戶(hù)并生成ssh key su dscheduler; ssh-keygen -t rsa; ssh-copy-id -i ~/.ssh/id_rsa.pub dscheduler@[hostname];

創(chuàng)建ds的元數(shù)據(jù)庫(kù)-MySQL

CREATE DATABASE dolphinscheduler DEFAULT CHARACTER SET utf8 DEFAULT COLLATE utf8_general_ci; CREATE USER 'dscheduler'@'%' IDENTIFIED BY 'dscheduler'; GRANT ALL PRIVILEGES ON dolphinscheduler.* TO 'dscheduler'@'%' IDENTIFIED BY 'dscheduler'; flush privileges;

安裝包解壓&權(quán)限修改

上傳安裝包到集群/opt目錄，執(zhí)行解壓

# 解壓安裝包 tar -zxvf apache-dolphinscheduler-incubating-1.3.1-cdh5.16.2-dolphinscheduler-bin.tar.gz -C /opt/# 重命名 mv apache-dolphinscheduler-incubating-1.3.1-cdh5.16.2-dolphinscheduler-bin ds-1.3.1-cdh5.16.2# 修改文件權(quán)限和組 chmod -R 755 ds-1.3.1-cdh5.16.2; chown -R dscheduler:dscheduler ds-1.3.1-cdh5.16.2;

初始化MySQL

修改數(shù)據(jù)庫(kù)配置

這里特別注意數(shù)據(jù)庫(kù)的連接配置

這里特別注意數(shù)據(jù)庫(kù)的連接配置

這里特別注意數(shù)據(jù)庫(kù)的連接配置

vi /opt/ds-1.3.1-cdh5.16.2/conf/datasource.properties;spring.datasource.driver-class-name=com.mysql.jdbc.Driver spring.datasource.url=jdbc:mysql://cm.eights.com:3306/dolphinscheduler?useUnicode=true&characterEncoding=UTF-8&allowMultiQueries=true spring.datasource.username=dscheduler spring.datasource.password=dscheduler

在ds的安裝包目錄下執(zhí)行數(shù)據(jù)庫(kù)初始化腳本

./script/create-dolphinscheduler.sh

配置ds所需的環(huán)境變量，這里記住一定要配，ds在執(zhí)行任務(wù)的時(shí)候會(huì)先source dolphinscheduler_env.sh

vi /opt/ds-1.3.1-cdh5.16.2/conf/env/dolphinscheduler_env.sh# 測(cè)試集群上沒(méi)有datax和flink請(qǐng)忽略相關(guān)配置 export HADOOP_HOME=/opt/cloudera/parcels/CDH/lib/hadoop export HADOOP_CONF_DIR=/opt/cloudera/parcels/CDH/lib/hadoop/etc/hadoop export SPARK_HOME1=/opt/cloudera/parcels/CDH/lib/spark export SPARK_HOME2=/opt/cloudera/parcels/SPARK2/lib/spark2 export PYTHON_HOME=/usr/local/anaconda3/bin/python export JAVA_HOME=/usr/java/jdk1.8.0_131 export HIVE_HOME=/opt/cloudera/parcels/CDH/lib/hive export FLINK_HOME=/opt/soft/flink export DATAX_HOME=/opt/soft/datax/bin/datax.py

編寫(xiě)ds的配置文件

ds在1.3.0之前，一鍵部署的配置文件在install.sh中。1.3.0版本的install.sh腳本只是一個(gè)部署腳本，部署配置文件在conf/config/install_config.conf中。install_config.conf精簡(jiǎn)了很多非必須的參數(shù)，如果要進(jìn)一步的進(jìn)行參數(shù)調(diào)整，則需要去修改conf下對(duì)應(yīng)模塊的配置文件。

下面給出本次部署的集群部署配置

# NOTICE : If the following config has special characters in the variable `.*[]^${}\+?|()@#&`, Please escape, for example, `[` escape to `\[` # postgresql or mysql dbtype="mysql"# db config # db address and port dbhost="cm.eights.com:3306"# db username username="dscheduler"# database name dbname="dolphinscheduler"# db passwprd # NOTICE: if there are special characters, please use the \ to escape, for example, `[` escape to `\[` password="dscheduler"# zk cluster zkQuorum="master.eights.com:2181,dn1.eights.com:2181,cm.eights.com:2181"# Note: the target installation path for dolphinscheduler, please not config as the same as the current path (pwd) # 在每臺(tái)部署機(jī)器上 安裝ds的目錄，角色的日志文件和任務(wù)日志都在這個(gè)目錄下 installPath="/opt/ds-1.3.1-agent"# deployment user # Note: the deployment user needs to have sudo privileges and permissions to operate hdfs. If hdfs is enabled, the root directory needs to be created by itself deployUser="dscheduler"# 郵件我這邊是內(nèi)網(wǎng)郵箱，后續(xù)會(huì)出外網(wǎng)郵件的配置方法 # alert config # mail server host mailServerHost="xxxx"# mail server port # note: Different protocols and encryption methods correspond to different ports, when SSL/TLS is enabled, make sure the port is correct. mailServerPort="25"# sender mailSender="xxxx"# user mailUser="xxxx"# sender password # note: The mail.passwd is email service authorization code, not the email login password. mailPassword="xxxx"# TLS mail protocol support starttlsEnable="false"# SSL mail protocol support # only one of TLS and SSL can be in the true state. sslEnable="false"#note: sslTrust is the same as mailServerHost sslTrust="xxxxxx"# resource storage type：HDFS,S3,NONE resourceStorageType="HDFS"# 單點(diǎn)的HDFS和yarn直接進(jìn)行配置即可 # if resourceStorageType is HDFS，defaultFS write namenode address，HA you need to put core-site.xml and hdfs-site.xml in the conf directory. # if S3，write S3 address，HA，for example ：s3a://dolphinscheduler， # Note，s3 be sure to create the root directory /dolphinscheduler defaultFS="hdfs://master.eights.com:8020"# if resourceStorageType is S3, the following three configuration is required, otherwise please ignore s3Endpoint="http://192.168.xx.xx:9010" s3AccessKey="xxxxxxxxxx" s3SecretKey="xxxxxxxxxx"# 這里注意，即使yarn是單點(diǎn)也最好把yarnHaIps配上 # if resourcemanager HA enable, please type the HA ips ; if resourcemanager is single, make this value empty yarnHaIps="master.eights.com"# if resourcemanager HA enable or not use resourcemanager, please skip this value setting; If resourcemanager is single, you only need to replace yarnIp1 to actual resourcemanager hostname. singleYarnIp="master.eights.com"# resource store on HDFS/S3 path, resource file will store to this hadoop hdfs path, self configuration, please make sure the directory exists on hdfs and have read write permissions。/dolphinscheduler is recommended resourceUploadPath="/dolphinscheduler"# who have permissions to create directory under HDFS/S3 root path # Note: if kerberos is enabled, please config hdfsRootUser= hdfsRootUser="hdfs"# kerberos config # whether kerberos starts, if kerberos starts, following four items need to config, otherwise please ignore kerberosStartUp="false" # kdc krb5 config file path krb5ConfPath="$installPath/conf/krb5.conf" # keytab username keytabUserName="hdfs-mycluster@ESZ.COM" # username keytab path keytabPath="$installPath/conf/hdfs.headless.keytab"# api server port apiServerPort="12345"# install hosts # Note: install the scheduled hostname list. If it is pseudo-distributed, just write a pseudo-distributed hostname ips="master.eights.com,dn1.eights.com,dn2.eights.com"# ssh port, default 22 # Note: if ssh port is not default, modify here sshPort="22"# run master machine # Note: list of hosts hostname for deploying master masters="master.eights.com,dn1.eights.com"# run worker machine # 1.3.1把worker分組從mysql移到了zookeeper中，這個(gè)需要打worker的標(biāo)簽 # 目前還不支持一個(gè)worker屬于多個(gè)分組 # note: need to write the worker group name of each worker, the default value is "default" workers="master.eights.com:default,dn2.eights.com:sqoop"# run alert machine # note: list of machine hostnames for deploying alert server alertServer="dn1.eights.com"# run api machine # note: list of machine hostnames for deploying api server apiServers="dn1.eights.com"

添加Hadoop集群配置文件

• 如果集群未啟用HA，直接在install_config.conf文件中進(jìn)行編寫(xiě)
• 如果集群?jiǎn)⒂昧薍A，請(qǐng)將hadoop的hdfs-site.xml和core-site.xml拷貝到/conf目錄下

修改JVM參數(shù)

• 兩個(gè)文件
  • /bin/dolphinscheduler-daemon.sh
  • /scripts/dolphinscheduler-daemon.sh

export DOLPHINSCHEDULER_OPTS="-server -Xmx16g -Xms1g -Xss512k -XX:+DisableExplicitGC -XX:+UseConcMarkSweepGC -XX:+CMSParallelRemarkEnabled -XX:LargePageSizeInBytes=128m -XX:+UseFastAccessorMethods -XX:+UseCMSInitiatingOccupancyOnly -XX:CMSInitiatingOccupancyFraction=70"

一鍵部署&進(jìn)程檢查&單服務(wù)啟停

# 運(yùn)行部署ds腳本 sh install.sh# 進(jìn)程檢查 jps

服務(wù)啟停

# 一鍵停止 sh ./bin/stop-all.sh # 一鍵開(kāi)啟 sh ./bin/start-all.sh # 啟停master sh ./bin/dolphinscheduler-daemon.sh start master-server sh ./bin/dolphinscheduler-daemon.sh stop master-server # 啟停worker sh ./bin/dolphinscheduler-daemon.sh start worker-server sh ./bin/dolphinscheduler-daemon.sh stop worker-server # 啟停api-server sh ./bin/dolphinscheduler-daemon.sh start api-server sh ./bin/dolphinscheduler-daemon.sh stop api-server # 啟停logger sh ./bin/dolphinscheduler-daemon.sh start logger-server sh ./bin/dolphinscheduler-daemon.sh stop logger-server # 啟停alert sh ./bin/dolphinscheduler-daemon.sh start alert-server sh ./bin/dolphinscheduler-daemon.sh stop alert-server

前端訪問(wèn)

dolphinscheduler-1.3.1前端不在需要nginx，直接使用

apiserver:12345/dolphinscheduler進(jìn)行訪問(wèn)

賬號(hào) admin 密碼 dolphinscheduler123登陸

檢查worker分組

• 可以看到1.3.1版本的worker分組是通過(guò)install_config.conf去執(zhí)行的，在頁(yè)面上不能進(jìn)行修改

檢查服務(wù)

編譯好的Dolphin Scheduler-1.3.1-cdh5.16.2的包

百度網(wǎng)盤(pán)

鏈接：

https://pan.baidu.com/s/1gEwEF2R2XJVRv76SgiW0hA

提取碼：joyq

6

總結(jié)

dolphinscheduler-1.3.1版本在部署上大幅精簡(jiǎn)了install.sh的配置，可以讓用戶(hù)快速部署起來(lái)。不過(guò)如果做升級(jí)的老鐵就需要去conf目錄下修改對(duì)應(yīng)模塊的配置文件。

前端這塊也不再需要單獨(dú)的nginx，用戶(hù)的部署體驗(yàn)更佳

總結(jié)

以上是生活随笔為你收集整理的如何在CDH5上部署Dolphin Scheduler 1.3.1的全部?jī)?nèi)容，希望文章能夠幫你解決所遇到的問(wèn)題。

如果覺(jué)得生活随笔網(wǎng)站內(nèi)容還不錯(cuò)，歡迎將生活随笔推薦給好友。