hadoop - 运行 oozie fork 时出现 LeaseExpiredException

Question

我们正在尝试运行一个 Oozie 工作流，其中 3 个子工作流使用 fork 并行运行。子工作流包含一个运行原生 map reduce 作业的节点，以及随后运行一些复杂的 PIG 作业的两个节点。最后，三个子工作流被连接到一个 end 节点。

当我们运行这个工作流时，我们得到 LeaseExpiredException。异常在运行 PIG 作业时随机发生。它发生的时间没有确定的位置，但是每次我们运行 WF 时都会发生。

此外，如果我们删除 fork 并按顺序运行子工作流，它会正常工作。然而，我们的期望是让它们在某些执行时间内并行运行。

能否请您帮助我理解这个问题以及一些关于我们可能出错的地方的指示。我们从 hadoop 开发开始，以前从未遇到过这样的问题。

看起来由于多个任务并行运行，其中一个线程关闭了一个零件文件，当另一个线程试图关闭该文件时，它抛出了错误。

以下是 hadoop 日志中异常的堆栈跟踪。

2013-02-19 10:23:54,815 INFO org.apache.pig.backend.hadoop.executionengine.mapReduceLayer.MapReduceLauncher: 57% complete 
2013-02-19 10:26:55,361 INFO org.apache.pig.backend.hadoop.executionengine.mapReduceLayer.MapReduceLauncher: 59% complete
2013-02-19 10:27:59,666 ERROR org.apache.hadoop.hdfs.DFSClient: Exception closing file <hdfspath>/oozie-oozi/0000105-130218000850190-oozie-oozi-W/aggregateData--pig/output/_temporary/_attempt_201302180007_0380_m_000000_0/part-00000 : org.apache.hadoop.ipc.RemoteException: org.apache.hadoop.hdfs.server.namenode.LeaseExpiredException: No lease on <hdfspath>/oozie-oozi/0000105-130218000850190-oozie-oozi-W/aggregateData--pig/output/_temporary/_attempt_201302180007_0380_m_000000_0/part-00000 File does not exist. Holder DFSClient_attempt_201302180007_0380_m_000000_0 does not have any open files.
                at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.checkLease(FSNamesystem.java:1664)
                at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.checkLease(FSNamesystem.java:1655)
                at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.completeFileInternal(FSNamesystem.java:1710)
                at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.completeFile(FSNamesystem.java:1698)
                at org.apache.hadoop.hdfs.server.namenode.NameNode.complete(NameNode.java:793)
                at sun.reflect.GeneratedMethodAccessor34.invoke(Unknown Source)
                at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)
                at java.lang.reflect.Method.invoke(Method.java:597)
                at org.apache.hadoop.ipc.RPC$Server.call(RPC.java:557)
                at org.apache.hadoop.ipc.Server$Handler$1.run(Server.java:1439)
                at org.apache.hadoop.ipc.Server$Handler$1.run(Server.java:1435)
                at java.security.AccessController.doPrivileged(Native Method)
                at javax.security.auth.Subject.doAs(Subject.java:396)
                at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1278)
                at org.apache.hadoop.ipc.Server$Handler.run(Server.java:1433)

以下是主工作流程和一个子工作流程的示例。

主要工作流程:

<workflow-app xmlns="uri:oozie:workflow:0.2" name="MainProcess">
<start to="forkProcessMain"/>
<fork name="forkProcessMain">
    <path start="Proc1"/>
    <path start="Proc2"/>
    <path start="Proc3"/>
</fork>
<join name="joinProcessMain" to="end"/>
<action name="Proc1">
    <sub-workflow>
        <app-path>${nameNode}${wfPath}/proc1_workflow.xml</app-path>
        <propagate-configuration/>
    </sub-workflow>
    <ok to="joinProcessMain"/>
    <error to="fail"/>
</action>   
<action name="Proc2">
    <sub-workflow>
        <app-path>${nameNode}${wfPath}/proc2_workflow.xml</app-path>
        <propagate-configuration/>
    </sub-workflow>
    <ok to="joinProcessMain"/>
    <error to="fail"/>
</action>   
<action name="Proc3">
    <sub-workflow>
        <app-path>${nameNode}${wfPath}/proc3_workflow.xml</app-path>
        <propagate-configuration/>
    </sub-workflow>
    <ok to="joinProcessMain"/>
    <error to="fail"/>
</action>   
<kill name="fail">
    <message>WF Failure, 'wf:lastErrorNode()' failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message>
</kill>
<end name="end"/>

子工作流程:

<workflow-app xmlns="uri:oozie:workflow:0.2" name="Sub Process">
<start to="Step1"/>
<action name="Step1">
    <java>
        <job-tracker>${jobTracker}</job-tracker>
        <name-node>${nameNode}</name-node>
        <prepare>
           <delete path="${step1JoinOutputPath}"/>
        </prepare>
        <configuration>
            <property>
                <name>mapred.queue.name</name>
                <value>${queueName}</value>
            </property>
        </configuration>
        <main-class>com.absd.mr.step1</main-class>
        <arg>${wf:name()}</arg>
        <arg>${wf:id()}</arg>
        <arg>${tbMasterDataOutputPath}</arg>
        <arg>${step1JoinOutputPath}</arg>
        <arg>${tbQueryKeyPath}</arg>
        <capture-output/>
    </java>
    <ok to="generateValidQueryKeys"/>
    <error to="fail"/>
</action>
<action name="generateValidQueryKeys">
    <pig>
        <job-tracker>${jobTracker}</job-tracker>
        <name-node>${nameNode}</name-node>
        <prepare>
           <delete path="${tbValidQuerysOutputPath}"/>
        </prepare>
        <configuration>
            <property>
                <name>pig.tmpfilecompression</name>
                <value>true</value>
            </property>
            <property>
                <name>pig.tmpfilecompression.codec</name>
                <value>lzo</value>
            </property>
            <property>
                <name>pig.output.map.compression</name>
                <value>true</value>
            </property>
            <property>
                <name>pig.output.map.compression.codec</name>
                <value>lzo</value>
            </property>
            <property>
                <name>pig.output.compression</name>
                <value>true</value>
            </property>
            <property>
                <name>pig.output.compression.codec</name>
                <value>lzo</value>
            </property>
            <property>
                <name>mapred.compress.map.output</name>
                <value>true</value>
            </property>
        </configuration>
        <script>${pigDir}/tb_calc_valid_accounts.pig</script>
        <param>csvFilesDir=${csvFilesDir}</param>
        <param>step1JoinOutputPath=${step1JoinOutputPath}</param>
        <param>tbValidQuerysOutputPath=${tbValidQuerysOutputPath}</param>
        <param>piMinFAs=${piMinFAs}</param>
        <param>piMinAccounts=${piMinAccounts}</param>
        <param>parallel=80</param>
    </pig>
    <ok to="aggregateAumData"/>
    <error to="fail"/>
</action>
<action name="aggregateAumData">
    <pig>
        <job-tracker>${jobTracker}</job-tracker>
        <name-node>${nameNode}</name-node>
        <prepare>
           <delete path="${tbCacheDataPath}"/>
        </prepare>
        <configuration>
            <property>
                <name>pig.tmpfilecompression</name>
                <value>true</value>
            </property>
            <property>
                <name>pig.tmpfilecompression.codec</name>
                <value>lzo</value>
            </property>
            <property>
                <name>pig.output.map.compression</name>
                <value>true</value>
            </property>
            <property>
                <name>pig.output.map.compression.codec</name>
                <value>lzo</value>
            </property>
            <property>
                <name>pig.output.compression</name>
                <value>true</value>
            </property>
            <property>
                <name>pig.output.compression.codec</name>
                <value>lzo</value>
            </property>
            <property>
                <name>mapred.compress.map.output</name>
                <value>true</value>
            </property>
        </configuration>
        <script>${pigDir}/aggregationLogic.pig</script>
        <param>csvFilesDir=${csvFilesDir}</param>
        <param>tbValidQuerysOutputPath=${tbValidQuerysOutputPath}</param>
        <param>tbCacheDataPath=${tbCacheDataPath}</param>
        <param>currDate=${date}</param>
        <param>udfJarPath=${nameNode}${wfPath}/lib</param>
        <param>parallel=150</param>
      </pig>
    <ok to="loadDataToDB"/>
    <error to="fail"/>
</action>   
<kill name="fail">
    <message>WF Failure, 'wf:lastErrorNode()' failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message>
</kill>
<end name="end"/>

Answer 1

当我们同时运行三个 pig Action 并且其中一个失败时，我们遇到了同样的错误。该消息错误是工作流意外停止的结果，因为一个操作失败，工作流已停止，而其他操作正在尝试继续。 您必须查看状态为 ERROR 的失败操作才能知道发生了什么，不要查看状态为 KILLED 的操作