oozie 入门

转自：http://blackproof.iteye.com/blog/1928122

oozie概述：oozie能干什么

oozie格式：怎么用oozie

oozie执行：怎么运行oozie

oozie概述：

oozie是基于hadoop的调度器，以xml的形式写调度流程，可以调度mr，pig，hive，shell，jar等等。

主要的功能有

Workflow： 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，定时触发workflow

Bundle Job，绑定多个coordinator

oozie格式：

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

一、job.properties里定义环境变量

nameNode	hdfs://xxx5:8020	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

注意：

workflow：oozie.wf.application.path

coordinator：oozie.coord.application.path

bundle：oozie.bundle.application.path

二、XML

1.workflow：

1. <workflow-app xmlns="uri:oozie:workflow:0.2" name="wf-example1">
2. <start to="pig-node">
3. <action name="pig-node">
4. <pig>
5. <job-tracker>${jobTracker}</job-tracker>
6. <name-node>${nameNode}</name-node>
7. <prepare>
8. <delete path="hdfs://xxx5/user/hadoop/appresult" />
9. </prepare>
10. <configuration>
11. <property>
12. <name>mapred.job.queue.name</name>
13. <value>default</value>
14. <property>
15. <property>
16. <name>mapred.compress.map.output</name>
17. <value>true</value>
18. <property>
19. <property>
20. <name>mapreduce.fileoutputcommitter.marksuccessfuljobs</name>
21. <value>false</value>
22. <property>
23. </configuration>
24. <script>test.pig</script>
25. <param>filepath=${filpath}</param>
26. </pig>
27. <ok to="end">
28. <error to="fail">
29. </action>
30. <kill name="fail">
31. <message>
32. Map/Reduce failed, error               message[${wf:errorMessage(wf:lastErrorNode())}]
33. </message>
34. </kill>
35. <end name="end"/>
36. </workflow-app>

2.coordinator

1. <coordinator-app name="cron-coord" frequence="${coord:hours(6)}" start="${start}" end="${end}"
2. timezoe="UTC" xmlns="uri:oozie:coordinator:0.2">
3. <action>
4. <workflow>
5. <app-path>${nameNode}/user/{$coord:user()}/${examplesRoot}/wpath</app-path>
6. <configuration>
7. <property>
8. <name>jobTracker</name>
9. <value>${jobTracker}</value>
10. </property>
11. <property>
12. <name>nameNode</name>
13. <value>${nameNode}</value>
14. </property>
15. <property>
16. <name>queueName</name>
17. <value>${queueName}</value>
18. </property>
19. </configuration>
20. </workflow>
21. </action>

注意：coordinator设置的UTC，比北京时间晚8个小时，所以你要是把期望执行时间减8小时

coordinator里面传值给workflow，example，时间设置为亚洲

1. <coordinator-app name="gwk-hour-log-coord" frequency="${coord:hours(1)}" start="${hourStart}" end="${hourEnd}" timezone="Asia/Shanghai"
2. xmlns="uri:oozie:coordinator:0.2">
3. <action>
4. <workflow>
5. <app-path>${workflowHourLogAppUri}/gwk-workflow.xml</app-path>
6. <configuration>
7. <property>
8. <name>yyyymmddhh</name>
9. <value>${coord:formatTime(coord:dateOffset(coord:nominalTime(),-1,'HOUR'), 'yyyyMMddHH')}</value>
10. </property>
11. </configuration>
12. </workflow>
13. </action>
14. </coordinator-app>

3.bundle

1. <bundle-app name='APPNAME' xmlns:xsi='http://www.w3.org/2001/XMLSchema-instance' xmlns='uri:oozie:bundle:0.1'>
2. <controls>
3. <kick-off-time>${kickOffTime}</kick-off-time>
4. </controls>
5. <coordinator name='coordJobFromBundle1' >
6. <app-path>${appPath}</app-path>
7. <configuration>
8. <property>
9. <name>startTime1</name>
10. <value>${START_TIME}</value>
11. </property>
12. <property>
13. <name>endTime1</name>
14. <value>${END_TIME}</value>
15. </property>
16. </configuration>
17. </coordinator>
18. <coordinator name='coordJobFromBundle2' >
19. <app-path>${appPath2}</app-path>
20. <configuration>
21. <property>
22. <name>startTime2</name>
23. <value>${START_TIME2}</value>
24. </property>
25. <property>
26. <name>endTime2</name>
27. <value>${END_TIME2}</value>
28. </property>
29. </configuration>
30. </coordinator>
31. </bundle-app>

oozie hive

1. <action name="hive-app">
2. <hive xmlns="uri:oozie:hive-action:0.2">
3. <job-tracker>${jobTracker}</job-tracker>
4. <name-node>${nameNode}</name-node>
5. <job-xml>hive-site.xml</job-xml>
6. <script>hivescript.q</script>
7. <param>yyyymmdd=${yyyymmdd}</param>
8. <param>yesterday=${yesterday}</param>
9. <param>lastmonth=${lastmonth}</param>
10. </hive>
11. <ok to="result-stat-join"/>
12. <error to="fail"/>
13. </action>

oozie运行

启动任务：

1. oozie job -oozie http://xxx5:11000/oozie -config job.properties -run

停止任务：

oozie job -oozie http://localhost:8080/oozie -kill 14-20090525161321-oozie-joe

注意：在停止任务的时候，有的时候会出现全线问题，需要修改oozie-site.xml文件

hadoop.proxyuser.oozie.groups *

hadoop.proxyuser.oozie.hosts *

oozie.server.ProxyUserServer.proxyuser.hadoop.hosts *

oozie.server.ProxyUserServer.proxyuser.hadoop.groups *