Hadoop运行MapReduce任务时，异常及解决方案

异常代码描述

刚刚开始接触Hadoop，对于MapReduce并不时特别了解，以下记录以下纠结了一天的问题及解决方案

1、执行MapReduce任务

hadoop jar wc.jar hejie.zheng.mapreduce.wordcount2.WordCountDriver /input /output

2、跳出异常 Task failed task_1643869122334_0004_m_000000

[2022-02-03 15:10:59.255]Container killed on request. Exit code is 143
[2022-02-03 15:10:59.333]Container exited with a non-zero exit code 143. 

2022-02-03 15:11:11,860 INFO mapreduce.Job: Task Id : attempt_1643869122334_0004_m_000000_2, Status : FAILED
[2022-02-03 15:11:10.394]Container [pid=4334,containerID=container_1643869122334_0004_01_000004] is running 273537536B beyond the 'VIRTUAL' memory limit. Current usage: 68.0 MB of 1 GB physical memory used; 2.4 GB of 2.1 GB virtual memory used. Killing container.
Dump of the process-tree for container_1643869122334_0004_01_000004 :
	|- PID PPID PGRPID SESSID CMD_NAME USER_MODE_TIME(MILLIS) SYSTEM_TIME(MILLIS) VMEM_USAGE(BYTES) RSSMEM_USAGE(PAGES) FULL_CMD_LINE
	|- 4343 4334 4334 4334 (java) 543 250 2518933504 17149 /opt/module/jdk1.8.0_212/bin/java -Djava.net.preferIPv4Stack=true -Dhadoop.metrics.log.level=WARN -Xmx820m -Djava.io.tmpdir=/opt/module/hadoop-3.1.3/data/nm-local-dir/usercache/user01/appcache/application_1643869122334_0004/container_1643869122334_0004_01_000004/tmp -Dlog4j.configuration=container-log4j.properties -Dyarn.app.container.log.dir=/opt/module/hadoop-3.1.3/logs/userlogs/application_1643869122334_0004/container_1643869122334_0004_01_000004 -Dyarn.app.container.log.filesize=0 -Dhadoop.root.logger=INFO,CLA -Dhadoop.root.logfile=syslog org.apache.hadoop.mapred.YarnChild 192.168.10.102 39922 attempt_1643869122334_0004_m_000000_2 4 
	|- 4334 4332 4334 4334 (bash) 2 6 9461760 270 /bin/bash -c /opt/module/jdk1.8.0_212/bin/java -Djava.net.preferIPv4Stack=true -Dhadoop.metrics.log.level=WARN   -Xmx820m -Djava.io.tmpdir=/opt/module/hadoop-3.1.3/data/nm-local-dir/usercache/user01/appcache/application_1643869122334_0004/container_1643869122334_0004_01_000004/tmp -Dlog4j.configuration=container-log4j.properties -Dyarn.app.container.log.dir=/opt/module/hadoop-3.1.3/logs/userlogs/application_1643869122334_0004/container_1643869122334_0004_01_000004 -Dyarn.app.container.log.filesize=0 -Dhadoop.root.logger=INFO,CLA -Dhadoop.root.logfile=syslog org.apache.hadoop.mapred.YarnChild 192.168.10.102 39922 attempt_1643869122334_0004_m_000000_2 4 1>/opt/module/hadoop-3.1.3/logs/userlogs/application_1643869122334_0004/container_1643869122334_0004_01_000004/stdout 2>/opt/module/hadoop-3.1.3/logs/userlogs/application_1643869122334_0004/container_1643869122334_0004_01_000004/stderr  

[2022-02-03 15:11:10.532]Container killed on request. Exit code is 143
[2022-02-03 15:11:10.562]Container exited with a non-zero exit code 143. 

2022-02-03 15:11:22,338 INFO mapreduce.Job:  map 100% reduce 100%
2022-02-03 15:11:23,386 INFO mapreduce.Job: Job job_1643869122334_0004 failed with state FAILED due to: Task failed task_1643869122334_0004_m_000000
Job failed as tasks failed. failedMaps:1 failedReduces:0 killedMaps:0 killedReduces: 0

2022-02-03 15:11:23,653 INFO mapreduce.Job: Counters: 13
	Job Counters 
		Failed map tasks=4
		Killed reduce tasks=1
		Launched map tasks=4
		Other local map tasks=3
		Data-local map tasks=1
		Total time spent by all maps in occupied slots (ms)=30921
		Total time spent by all reduces in occupied slots (ms)=0
		Total time spent by all map tasks (ms)=30921
		Total vcore-milliseconds taken by all map tasks=30921
		Total megabyte-milliseconds taken by all map tasks=31663104
	Map-Reduce Framework
		CPU time spent (ms)=0
		Physical memory (bytes) snapshot=0
		Virtual memory (bytes) snapshot=0

3、解决方案

先上解决方法，在描述问题原因
参考文章中的配置。
首先，跳转到hadoop-3.1.3\etc\hadoop 目录下，执行：

vim mapred-site.xml

添加如下代码，配置运行时内存（如文件内容比较大，可以根据实践情况配置）

<property>
    <name>mapreduce.map.memory.mb</name>
    <value>1024</value>
    <description>修改每个Map task容器申请的内存大小；默认1G</description>
</property>
<property>
    <name>mapreduce.map.java.opts</name>
    <value>-Xmx64m</value>
</property>
<property>
    <name>mapreduce.reduce.memory.mb</name>
    <value>1024</value>
</property>
<property>
    <name>mapreduce.reduce.java.opts</name>
    <value>-Xmx64m</value>
    <description>一般设置为mapreduce.map.memory.mb的85%左右</description>
</property>
<property>
    <name>mapreduce.task.io.sort.mb</name>
    <value>4</value>

4、刨根问底

其实可以通过http://hadoop103:8088/cluster，看出当前任务的history，查看具体的问题出现在哪里：

可以看到MapReduce申请了2.4GB的内存空间，超过了默认的虚拟内存2.1GB，内存不够导致异常。
修改mapred-site.xml内存大小即可。
（1）mapreduce.map.memory.mb 默认1G，map执行时分配的内存

（2）mapreduce.reduce.memory.mb 默认1G，reduce执行时分配的内存

（3）yarn.app.mapreduce.am.resource.mb 默认1.5G， MR AppMaster分配的内存
jvm参数默认值
有下面的一些jvm的参数，他们的默认值一般比较容易忽略，记录下来，方便查询
-Xms 默认情况下堆内存的64分之一，最小1M
-Xmx 默认情况下堆内存的4分之一或1G
-Xmn 默认情况下堆内存的64分之一
-XX:NewRatio 默认为2
-XX:SurvivorRatio 默认为8

o -Xms:初始堆大小，如4096M
o -Xmx:最大堆大小，如4096M,如果xms>xmx, 那默认空余堆内存小于40%会触发调整到xmx，但是MinheapFreeRation可以调整
o -XX:NewSize=n:设置年轻代大小，如1024
o -Xmn1024M 设置年轻代大小为1024M
o -XX:NewRatio=n:设置年老代和年轻代的比值。建议值为3至5，如4，表示年老代：年轻代=4:1，年轻代占整个年轻代年老代和的1/5
o -XX:SurvivorRatio=n:年轻代中Eden区与两个Survivor区的比值。建议值3-5。注意Survivor区有两个。如8，表示2*Survivor:Eden=2:8，一个Survivor区占整个年轻代的1/10

-Xmx64M：最大堆空间64MB

深入了解内存设置可参考Yarn下Mapreduce的内存参数理解&xml参数配置

以上参数可以根据实践需要进行调整以达到最优效果。