oozie说明(本文参考多处,自己留看)

Oozie概述:

  Oozie是一个基于Hadoop工作流引擎,也可以称为调度器,它以xml的形式写调度流程,可以调度mr,pig,hive,shell,jar,spark等等。在实际工作中,遇到对数据进行一连串的操作的时候很实用,不需要自己写一些处理代码了,只需要定义好各个action,然后把他们串在一个工作流里面就可以自动执行了。对于大数据的分析工作非常有用. (以下介绍内容基于Oozie 4.1.0版本)

 Oozie有几个主要概念:

  workflow :工作流 ,顺序执行流程节点,支持fork(分支多个节点),join(合并多个节点为一个)。

  coordinator :多个workflow可以组成一个coordinator,可以把前几个workflow的输出作为后一个workflow的输入,也可以定义workflow的触发条件,来做定时触发。

  bundle: 是对一堆coordinator的抽象, 可绑定多个coordinator。

  job.properties:定义环境变量。

oozie安装 略

生命周期:

在Oozie中,工作流的状态可能存在如下几种:

状态

含义说明

PREP

一个工作流Job第一次创建将处于PREP状态,表示工作流Job已经定义,但是没有运行。

RUNNING

当一个已经被创建的工作流Job开始执行的时候,就处于RUNNING状态。它不会达到结束状态,只能因为出错而结束,或者被挂起。

SUSPENDED

一个RUNNING状态的工作流Job会变成SUSPENDED状态,而且它会一直处于该状态,除非这个工作流Job被重新开始执行或者被杀死。

SUCCEEDED

当一个RUNNING状态的工作流Job到达了end节点,它就变成了SUCCEEDED最终完成状态。

KILLED

当一个工作流Job处于被创建后的状态,或者处于RUNNING、SUSPENDED状态时,被杀死,则工作流Job的状态变为KILLED状态。

FAILED

当一个工作流Job不可预期的错误失败而终止,就会变成FAILED状态。

上述各种状态存在相应的转移(工作流程因为某些事件,可能从一个状态跳转到另一个状态),其中合法的状态转移有如下几种,如下表所示:

转移前状态

转移后状态集合

未启动

PREP

PREP

RUNNING、KILLED

RUNNING

SUSPENDED、SUCCEEDED、KILLED、FAILED

SUSPENDED

RUNNING、KILLED

明确上述给出的状态转移空间以后,可以根据实际需要更加灵活地来控制工作流Job的运行。

 

oozie格式:

1.workflow:

 Oozie定义了一种基于XML的hPDL (Hadoop Process Definition Language)来描述workflow的DAG。在workflow中定义了控制流节点(Control Flow Nodes)、动作节点(Action Nodes)

其中,控制流节点定义了流程的开始和结束(start、end),以及控制流程的执行路径(Execution Path),如decision、fork、join等;而动作节点包括Hadoop任务、SSH、HTTP、eMail和Oozie子流程等。

Action Node定义了基本的工作任务节点。

语法:

 

<workflow-app name="[WF-DEF-NAME]" xmlns="uri:oozie:workflow:0.1">

  ...

    <start to="[NODE-NAME]"/>

   <action name="[NODE-NAME]">

          ....
     <ok to="[NODE-NAME]"/>

        <error to="[NODE-NAME]"/>

    </action> 
   <kill name="[NODE-NAME]"> <message>[MESSAGE-TO-LOG]</message> </kill>  

   <end name="[NODE-NAME]"/>

</workflow-app>

 

1.1 Map-Reduce Action

map-reduce动作会在工作流Job中启动一个MapReduce Job任务运行,我们可以详细配置这个MapReduce Job。另外,可以通过map-reduce元素的子元素来配置一些其他的任务,如streaming、pipes、file、archive等等。

语法:

<workflow-app name="[WF-DEF-NAME]" xmlns="uri:oozie:workflow:0.1">

    ...

    <action name="[NODE-NAME]">

        <map-reduce>

            <job-tracker>[JOB-TRACKER]</job-tracker>

            <name-node>[NAME-NODE]</name-node>

            <prepare>

                <delete path="[PATH]"/>

                ...

                <mkdir path="[PATH]"/>

                ...

            </prepare>

            <streaming>

                <mapper>[MAPPER-PROCESS]</mapper>

                <reducer>[REDUCER-PROCESS]</reducer>

                <record-reader>[RECORD-READER-CLASS]</record-reader>

                <record-reader-mapping>[NAME=VALUE]</record-reader-mapping>

                ...

                <env>[NAME=VALUE]</env>

                ...

            </streaming>

                                     <!-- Either streaming or pipes can be specified for an action, not both -->

            <pipes>

                <map>[MAPPER]</map>

                <reduce>[REDUCER]</reducer>

                <inputformat>[INPUTFORMAT]</inputformat>

                <partitioner>[PARTITIONER]</partitioner>

                <writer>[OUTPUTFORMAT]</writer>

                <program>[EXECUTABLE]</program>

            </pipes>

            <job-xml>[JOB-XML-FILE]</job-xml>

            <configuration>

                <property>

                    <name>[PROPERTY-NAME]</name>

                    <value>[PROPERTY-VALUE]</value>

                </property>

                ...

            </configuration>

            <file>[FILE-PATH]</file>

            ...

            <archive>[FILE-PATH]</archive>

            ...

        </map-reduce>        <ok to="[NODE-NAME]"/>

        <error to="[NODE-NAME]"/>

    </action>

    ...

</workflow-app>

官网给出的例子:

<workflow-app name="foo-wf" xmlns="uri:oozie:workflow:0.1">

    ...

    <action name="myfirstHadoopJob">

        <map-reduce>

            <job-tracker>foo:8021</job-tracker>

            <name-node>bar:8020</name-node>

            <prepare>

                <delete path="hdfs://foo:8020/usr/tucu/output-data"/>

            </prepare>

            <job-xml>/myfirstjob.xml</job-xml>

            <configuration>

                <property>

                    <name>mapred.input.dir</name>

                    <value>/usr/tucu/input-data</value>

                </property>

                <property>

                    <name>mapred.output.dir</name>

                    <value>/usr/tucu/input-data</value>

                </property>

                <property>

                    <name>mapred.reduce.tasks</name>

                    <value>${firstJobReducers}</value>

                </property>

                <property>

                    <name>oozie.action.external.stats.write</name>

                    <value>true</value>

                </property>

            </configuration>

        </map-reduce>

        <ok to="myNextAction"/>

        <error to="errorCleanup"/>

    </action>

    ...

</workflow-app>

 

1.2 Ssh Action

该动作主要是通过ssh登录到一台主机,能够执行一组shell命令.

注意: SSH actions在 Oozie schema 0.1中使用, 在Oozie schema 0.2已被删除.

       ssh action将一个shell命令作为一个远程安全的shell在远程主机后台启动. 工作流工作将等到远程shell命令完成后再继续下一个动作。shell命令必须存在于远程计算机中,必须通过命令路径执行它。

语法:

<workflow-app name="[WF-DEF-NAME]" xmlns="uri:oozie:workflow:0.1">

    ...

    <action name="[NODE-NAME]">

        <ssh>

            <host>[USER]@[HOST]</host>

            <command>[SHELL]</command>

            <args>[ARGUMENTS]</args>

            ...

            <capture-output/>

        </ssh>

        <ok to="[NODE-NAME]"/>

        <error to="[NODE-NAME]"/>

    </action>

    ...

</workflow-app>

 

官网给出的例子:

<workflow-app name="sample-wf" xmlns="uri:oozie:workflow:0.1">

    ...

    <action name="myssjob">

        <ssh>

            <host>foo@bar.com<host>

            <command>uploaddata</command>

            <args>jdbc:derby://bar.com:1527/myDB</args>

            <args>hdfs://foobar.com:8020/usr/tucu/myData</args>

        </ssh>

        <ok to="myotherjob"/>

        <error to="errorcleanup"/>

    </action>

    ...

</workflow-app>

 

 

1.3 Java Action

Oozie支持Java action ,Java action 会自动执行workflow任务中制定的java类中的 public static void main(String[] args)方法,会在hadoop集群上以单mapper task的形式执行一个map-reduce job.

 

workflow任务会等待当前java程序执行完继续执行下一个action,这意味着我们可以写多个action以此来调用多个类.  当java类正确执行退出后,将会进入ok控制流;当发生异常时,将会进入error控制流。

语法:

<workflow-app name="[WF-DEF-NAME]" xmlns="uri:oozie:workflow:0.1">

    ...

    <action name="[NODE-NAME]">

        <java>

            <job-tracker>[JOB-TRACKER]</job-tracker>

            <name-node>[NAME-NODE]</name-node>

            <prepare>

               <delete path="[PATH]"/>

               ...

               <mkdir path="[PATH]"/>

               ...

            </prepare>

            <job-xml>[JOB-XML]</job-xml>

            <configuration>

                <property>

                    <name>[PROPERTY-NAME]</name>

                    <value>[PROPERTY-VALUE]</value>

                </property>

                ...

            </configuration>

            <main-class>[MAIN-CLASS]</main-class>

                                     <java-opts>[JAVA-STARTUP-OPTS]</java-opts>

                                     <arg>ARGUMENT</arg>

            ...

            <file>[FILE-PATH]</file>

            ...

            <archive>[FILE-PATH]</archive>

            ...

            <capture-output />

        </java>

        <ok to="[NODE-NAME]"/>

        <error to="[NODE-NAME]"/>

    </action>

    ...

</workflow-app>

官网给出的例子:

<workflow-app name="sample-wf" xmlns="uri:oozie:workflow:0.1">

    ...

    <action name="myfirstjavajob">

        <java>

            <job-tracker>foo:8021</job-tracker>

            <name-node>bar:8020</name-node>

            <prepare>

                <delete path="${jobOutput}"/>

            </prepare>

            <configuration>

                <property>

                    <name>mapred.queue.name</name>

                    <value>default</value>

                </property>

            </configuration>

            <main-class>org.apache.oozie.MyFirstMainClass</main-class>

            <java-opts>-Dblah</java-opts>

                                     <arg>argument1</arg>

                                     <arg>argument2</arg>

        </java>

        <ok to="myotherjob"/>

        <error to="errorcleanup"/>

    </action>

    ...

</workflow-app>

1.4 shell action

Shell动作可以执行Shell命令,并通过配置命令所需要的参数。

语法:

<workflow-app name="[WF-DEF-NAME]" xmlns="uri:oozie:workflow:0.4">

 

 ...

 

 <action name="[NODE-NAME]">

 

     <shell xmlns="uri:oozie:shell-action:0.2">

 

         <job-tracker>[JOB-TRACKER]</job-tracker>

 

         <name-node>[NAME-NODE]</name-node>

 

         <prepare>

 

             <delete path="[PATH]" />

 

             ...

 

             <mkdir path="[PATH]" />

 

             ...

 

         </prepare>

 

         <configuration>

 

             <property>

 

                 <name>[PROPERTY-NAME]</name>

 

                 <value>[PROPERTY-VALUE]</value>

 

             </property>

 

             ...

 

         </configuration>

 

         <exec>[SHELL-COMMAND]</exec>

 

         <argument>[ARGUMENT-VALUE]</argument>

 

         <capture-output />

 

     </shell>

 

     <ok to="[NODE-NAME]" />

 

     <error to="[NODE-NAME]" />

 

</action>

 ...

 

</workflow-app>

1.5 Spark action

    Oozie支持Spark action,不过支持的不是特别好。提交spark任务时,需要加载spark-assembly jar。

语法:

<workflow-app name="[WF-DEF-NAME]" xmlns="uri:oozie:workflow:0.3">

    ...

    <action name="[NODE-NAME]">

        <spark xmlns="uri:oozie:spark-action:0.1">

            <job-tracker>[JOB-TRACKER]</job-tracker>

            <name-node>[NAME-NODE]</name-node>

            <prepare>

               <delete path="[PATH]"/>

               ...

               <mkdir path="[PATH]"/>

               ...

            </prepare>

            <job-xml>[SPARK SETTINGS FILE]</job-xml>

            <configuration>

                <property>

                    <name>[PROPERTY-NAME]</name>

                    <value>[PROPERTY-VALUE]</value>

                </property>

                ...

            </configuration>

            <master>[SPARK MASTER URL]</master>

            <mode>[SPARK MODE]</mode>

            <name>[SPARK JOB NAME]</name>

            <class>[SPARK MAIN CLASS]</class>

            <jar>[SPARK DEPENDENCIES JAR / PYTHON FILE]</jar>

            <spark-opts>[SPARK-OPTIONS]</spark-opts>

            <arg>[ARG-VALUE]</arg>

                ...

            <arg>[ARG-VALUE]</arg>

            ...

        </spark>

        <ok to="[NODE-NAME]"/>

        <error to="[NODE-NAME]"/>

    </action>

    ...

</workflow-app>

官网给的例子:

<workflow-app name="sample-wf" xmlns="uri:oozie:workflow:0.1">

    ...

    <action name="myfirstsparkjob">

        <spark xmlns="uri:oozie:spark-action:0.1">

            <job-tracker>foo:8021</job-tracker>

            <name-node>bar:8020</name-node>

            <prepare>

                <delete path="${jobOutput}"/>

            </prepare>

            <configuration>

                <property>

                    <name>mapred.compress.map.output</name>

                    <value>true</value>

                </property>

            </configuration>

            <master>local[*]</master>

            <mode>client</mode>

            <name>Spark Example</name>

            <class>org.apache.spark.examples.mllib.JavaALS</class>

            <jar>/lib/spark-examples_2.10-1.1.0.jar</jar>

            <spark-opts>--executor-memory 20G --num-executors 50

             --conf spark.executor.extraJavaOptions="-XX:+HeapDumpOnOutOfMemoryError -XX:HeapDumpPath=/tmp"</spark-opts>

            <arg>inputpath=hdfs://localhost/input/file.txt</arg>

            <arg>value=2</arg>

        </spark>

        <ok to="myotherjob"/>

        <error to="errorcleanup"/>

    </action>

    ...

</workflow-app>

2.coordinator.xml

语法:

 

<coordinator-app name="[NAME]" frequency="[FREQUENCY]"

                    start="[DATETIME]" end="[DATETIME]" timezone="[TIMEZONE]"

                    xmlns="uri:oozie:coordinator:0.1">   

#frequency:执行频率,小于五分钟要修改配置 start,end:开始与结束时间,若想跟北京时间一样也要修改配置文件,并修改时间格式

 

      <controls>

        <timeout>[TIME_PERIOD]</timeout>

        <concurrency>[CONCURRENCY]</concurrency>

        <execution>[EXECUTION_STRATEGY]</execution>

      </controls>

.

      <datasets>    

        <include>[SHARED_DATASETS]</include>

        ...

.

        <!-- Synchronous datasets --> #---数据生成目录

        <dataset name="[NAME]" frequency="[FREQUENCY]"

                 initial-instance="[DATETIME]" timezone="[TIMEZONE]">

          <uri-template>[URI_TEMPLATE]</uri-template>

        </dataset>

        ...

.

      </datasets>

.

      <input-events>    #----定义了数据触发条件

        <data-in name="[NAME]" dataset="[DATASET]">

          <instance>[INSTANCE]</instance>

          ...

        </data-in>

        ...

        <data-in name="[NAME]" dataset="[DATASET]">

          <start-instance>[INSTANCE]</start-instance>

          <end-instance>[INSTANCE]</end-instance>

        </data-in>

        ...

      </input-events>

      <output-events>

         <data-out name="[NAME]" dataset="[DATASET]">

           <instance>[INSTANCE]</instance>

         </data-out>

         ...

      </output-events>

      <action>

        <workflow>

          <app-path>[WF-APPLICATION-PATH]</app-path>    #---workflow.xml所在hdfs目录

          <configuration>

            <property>    #----定义传给workflow的参数

              <name>[PROPERTY-NAME]</name>

              <value>[PROPERTY-VALUE]</value>

            </property>

            ...

         </configuration>

       </workflow>

      </action>

   </coordinator-app>

 

官网给出的例子:

 

<coordinator-app name="hello-coord" frequency="${coord:days(1)}"

                    start="2009-01-02T08:00Z" end="2009-01-02T08:00Z"

                    timezone="America/Los_Angeles"

                    xmlns="uri:oozie:coordinator:0.1">

      <datasets>

        <dataset name="logs" frequency="${coord:days(1)}"

                 initial-instance="2009-01-02T08:00Z" timezone="America/Los_Angeles">

          <uri-template>hdfs://bar:8020/app/logs/${YEAR}${MONTH}/${DAY}/data</uri-template>

        </dataset>

        <dataset name="siteAccessStats" frequency="${coord:days(1)}"

                 initial-instance="2009-01-02T08:00Z" timezone="America/Los_Angeles">

          <uri-template>hdfs://bar:8020/app/stats/${YEAR}/${MONTH}/${DAY}/data</uri-template>

        </dataset>

      </datasets>

      <input-events>    

        <data-in name="input" dataset="logs">

          <instance>2009-01-02T08:00Z</instance>

        </data-in>

      </input-events>

      <output-events>

         <data-out name="output" dataset="siteAccessStats">

           <instance>2009-01-02T08:00Z</instance>

         </data-out>

      </output-events>

      <action>

        <workflow>

          <app-path>hdfs://bar:8020/usr/joe/logsprocessor-wf</app-path>   

          <configuration>

            <property>   

              <name>wfInput</name>

              <value>${coord:dataIn('input')}</value>

            </property>

            <property>

              <name>wfOutput</name>

              <value>${coord:dataOut('output')}</value>

            </property>

         </configuration>

       </workflow>

      </action>

   </coordinator-app>

 

 

3.bundle.xml

 

语法:

 

<bundle-app name=[NAME]  xmlns='uri:oozie:bundle:0.1'>

  <controls>

       <kick-off-time>[DATETIME]</kick-off-time>    #运行时间

  </controls>

   <coordinator name=[NAME] >

       <app-path>[COORD-APPLICATION-PATH]</app-path> # coordinator.xml所在目录

          <configuration>                 #传给coordinator应用的参数

            <property>

              <name>[PROPERTY-NAME]</name>  

              <value>[PROPERTY-VALUE]</value>

            </property>

            ...

         </configuration>

   </coordinator>

   ...

</bundle-app> 

 

官网给出的例子(绑定两个coordinator):

 

<bundle-app name='APPNAME' xmlns:xsi='http://www.w3.org/2001/XMLSchema-instance' xmlns='uri:oozie:bundle:0.1'>

  <controls>

       <kick-off-time>${kickOffTime}</kick-off-time>

  </controls>

   <coordinator name='coordJobFromBundle1' >

       <app-path>${appPath}</app-path>

       <configuration>

         <property>

              <name>startTime1</name>

              <value>${START_TIME}</value>

          </property>

         <property>

              <name>endTime1</name>

              <value>${END_TIME}</value>

          </property>

      </configuration>

   </coordinator>

   <coordinator name='coordJobFromBundle2' >

       <app-path>${appPath2}</app-path>

       <configuration>

         <property>

              <name>startTime2</name>

              <value>${START_TIME2}</value>

          </property>

         <property>

              <name>endTime2</name>

              <value>${END_TIME2}</value>

          </property>

      </configuration>

   </coordinator>

</bundle-app>

 

4,.job.properties:

 

nameNode               hdfs://xxx:8020    hdfs地址

jobTracker             xxx5:8034          jobTracker 地址

queueName              default            oozie队列

examplesRoot            examples           全局目录

oozie.usr.system.libpath    true           是否加载用户lib库

oozie.libpath            share/lib/user    用户lib库

oozie.wf.appication.path   ${nameNode}/user/${user.name}/... oozie流程所在hdfs地址

 

workflow:oozie.wf.application.path

coordinator:oozie.coord.application.path

bundle:oozie.bundle.application.path

Oozie使用:

写一个oozie,有两个是必要的:job.properties 和 workflow.xml(coordinator.xml,bundle.xml)

如果想让任务可以定时自动运行,那么需要写coordinator.xml。

如果想绑定多个coordinator.xml,那么需要写bundle.xml。

Oozie实例:

我们工作时的(简略版)实例:(本次以spark action为例)

bundle.xml:

 

<bundle-app name='APPNAME' xmlns:xsi='http://www.w3.org/2001/XMLSchema-instance'

xmlns='uri:oozie:bundle:0.2'>

    <coordinator name='coordJobFromBundle1' >

       <app-path>${appPath}</app-path>  

   </coordinator>

   <coordinator name='coordJobFromBundle2' >

       <app-path>${appPath2}</app-path>

   </coordinator>

 

</bundle-app>

 

coordinator.xml:

 

<coordinator-app name="cron-coord" frequency="${coord:minutes(6)}" start="${start}"

end="${end}" timezone="Asia/Shanghai" xmlns="uri:oozie:coordinator:0.2">

    <action>

        <workflow>

            <app-path>${workflowAppUri}</app-path>

            <configuration>

                <property>

                    <name>jobTracker</name>

                    <value>${jobTracker}</value>

                </property>

                <property>

                    <name>nameNode</name>

                    <value>${nameNode}</value>

                </property>

                <property>

                    <name>queueName</name>

                    <value>${queueName}</value>

                </property>

                <property>

                    <name>mainClass</name>

                    <value>com.ocn.itv.rinse.ErrorCollectRinse</value>

                </property>

                <property>

                    <name>mainClass2</name>

                    <value>com.ocn.itv.rinse.UserCollectRinse</value>

                </property>

                <property>

                    <name>jarName</name>

                    <value>ocn-itv-spark-3.0.3-rc1.jar</value>

                </property>

            </configuration>

        </workflow>

    </action>

</coordinator-app>

 

workflow.xml:

 

<workflow-app  name="spark-example1" xmlns="uri:oozie:workflow:0.5"> 

    <start to="forking"/>

    <fork name="forking">

        <path start="firstparalleljob"/>

        <path start="secondparalleljob"/>

    </fork>   

    <action name="firstparalleljob">

        <spark xmlns="uri:oozie:spark-action:0.2"> 

            <job-tracker>${jobTracker}</job-tracker> 

            <name-node>${nameNode}</name-node>

            <configuration> 

                <property> 

                    <name>mapred.job.queue.name</name> 

                    <value>${queueName}</value> 

                </property>                 

            </configuration>           

            <master>yarn-cluster</master>

            <mode>cluster</mode>

            <name>Spark Example</name>

            <class>${mainClass}</class>           

            <jar>${jarName}</jar>

            <spark-opts>${sparkopts}</spark-opts>

            <arg>${input}</arg>           

        </spark >  

        <ok to="joining"/>

        <error to="fail"/>   

    </action>

    <action name="secondparalleljob">

         <spark xmlns="uri:oozie:spark-action:0.2"> 

            <job-tracker>${jobTracker}</job-tracker> 

            <name-node>${nameNode}</name-node>

            <configuration> 

                <property> 

                    <name>mapred.job.queue.name</name> 

                    <value>${queueName}</value> 

                </property>                 

            </configuration>           

            <master>yarn-cluster</master>

            <mode>cluster</mode>

            <name>Spark Example2</name>

            <class>${mainClass2}</class>           

            <jar>${jarName}</jar>

            <spark-opts>${sparkopts}</spark-opts>

            <arg>${input}</arg>           

        </spark > 

        <ok to="joining"/>

        <error to="fail"/>   

    </action>  

    <join name="joining" to="end"/>

      <kill name="fail"> 

       <message>Action failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message> 

    </kill> 

   <end name="end"/> 

</workflow-app>

 

job.properties

 

nameNode=hdfs://hgdp-001:8020     #hsfs端口地址

jobTracker=hgdp-001:8032        #resourceManager的端口

queueName=default            #oozie队列

input=2017-05-09             #输入参数

hdfspath=user/root           #自定义目录

examplesRoot=ocn-itv-oozie      #自定义全局目录

oozie.use.system.libpath=True    #是否启动系统lib库

sparkopts=--executor-memory 1G    #参数设置

start=2017-09-04T00:05+0800    #coordinator任务开始时间

end=2017-09-04T00:36+0800      #coordinator任务结束时间

start2=2017-09-01T00:06+0800

end2=2017-09-04T00:36+0800

oozie.libpath=${nameNode}/${hdfspath}/${examplesRoot}/lib/          #用户自定义lib库(存放jar包)

workflowAppUri=${nameNode}/${hdfspath}/${examplesRoot}/wf/spark/fork/

workflowAppUri2=${nameNode}/${hdfspath}/${examplesRoot}/wf/spark/single/  #coordinator定时调度对应的workflow.xml所在目录

appPath=${nameNode}/${hdfspath}/${examplesRoot}/cd/single/

appPath2=${nameNode}/${hdfspath}/${examplesRoot}/cd/single1/        #bundle调用对应的coordinator.xml所在目录

oozie.bundle.application.path=${nameNode}/${hdfspath}/${examplesRoot}/bd/bd1/    #bundle.xml所在目录

#一个bundle调用多个coordinator

 

 

 

最后运行:

 

  启动任务:oozie job -config  job.properties  -run  -oozie http://192.168.2.11 (地址):11000/oozie

 

 

 需要注意的地方:

一.  coordinator中timezone的时区配置

Cloudera oozie默认时区是UTC,在开发oozie任务时必须在期望执行的时间上减去8小时,不方便。可以修改时区的配置操作。

1.在oozie的配置文件中添加如下属性:

<property>

 <name>oozie.processing.timezone</name>

 <value>GMT+0800</value>

</property>

2.如果使用了hue,进入Oozie web ui,选择Settings,然后在Timezone里选择CST(Asia/Shanghai)

3.coordinator中的timeone设置为:timezone="Asia/Shanghai"

4.修改时间格式,例如:2017-09-05T15:16+0800

二.oozie.xx.application.path

oozie.xx.application.path在job.properties里只能有一个。

workflow:oozie.wf.application.path

coordinator:oozie.coord.application.path

bundle:oozie.bundle.application.path

三.命名及存放位置问题

其中workflow.xml,coordinator.xml,bundle.xml名字都不可以修改,要放到hdfs目录中,而job.properties名字可以修改,放在本地即可。

四.关于workflow.xml 中action的问题:

可以写多个action依次执行,如下示例所示:

 

 

<workflow-app  name="java-example1" xmlns="uri:oozie:workflow:0.5"> 

    <start to="java-Action"/> 

    <action name="java-Action">

     ....

        <ok to="java-Action2"/>

        <error to="fail"/>   

    </action>

    <action name="java-Action2">

       ....

        <ok to="end"/>

        <error to="fail"/>   

    </action>  

      <kill name="fail"> 

       <message>Action failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message> 

    </kill> 

   <end name="end"/> 

</workflow-app>

 

 

也可以设置多个任务并发执行,需要添加fork和join节点,fork节点把任务切分成多个并行任务,join则合并多个并行任务。fork和join节点必须是成对出现的。join节点合并的任务,必须是通一个fork出来的子任务才行。示例如下:

 

<workflow-app  name="java-example1" xmlns="uri:oozie:workflow:0.5"> 

    <start to="forking"/>

    <fork name="forking">

        <path start="firstparalleljob"/>

        <path start="secondparalleljob"/>

    </fork>   

    <action name="firstparalleljob">

            .....

        <ok to="joining"/>

        <error to="fail"/>    

    </action>

    <action name="secondparalleljob">

            ....

        <ok to="joining"/>

        <error to="fail"/>   

    </action>  

    <join name="joining" to="end"/>

      <kill name="fail"> 

       <message>Action failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message> 

    </kill> 

   <end name="end"/> 

</workflow-app>

 

 

posted on 2017-09-14 14:07  风景1573  阅读(4181)  评论(0编辑  收藏  举报