项目场景:
环境: cdh集群
cdh版本: 6.3
问题描述
1.昨天下午收到同事反馈内网cdh集群有一台机器有问题
2.ssh一登录就报ERROR: Invalid HADOOP_MAPRED_HOME
3.输入任何大数据相关命令都报错:
ERROR: Invalid HADOOP_MAPRED_HOME
4.百度了一下,发现网上的解决方案都不奏效
解决方案:
1.看环境变量存在不
# echo $HADOOP_MAPRED_HOME
空的
2.查阅资料发现cdh 环境变量设置在 /etc/hadoop/conf/hadoop-env.sh
# cat /etc/hadoop/conf/hadoop-env.sh
内容如下:
# Prepend/Append plugin parcel classpaths
if [ "$HADOOP_USER_CLASSPATH_FIRST" = 'true' ]; then
# HADOOP_CLASSPATH={{HADOOP_CLASSPATH_APPEND}}
:
else
# HADOOP_CLASSPATH={{HADOOP_CLASSPATH}}
:
fi
# JAVA_LIBRARY_PATH={{JAVA_LIBRARY_PATH}}
export HADOOP_MAPRED_HOME=$( ([[ ! '/opt/cloudera/parcels/CDH/lib/hadoop-mapreduce' =~ CDH_MR2_HOME ]] && echo /opt/cloudera/parcels/CDH/lib/hadoop-mapreduce ) || echo ${CDH_MR2_HOME:-/usr/lib/hadoop-mapreduce/} )
export YARN_OPTS="-Xmx825955249 -Djava.net.preferIPv4Stack=true $YARN_OPTS"
export HADOOP_CLIENT_OPTS="-Djava.net.preferIPv4Stack=true $HADOOP_CLIENT_OPTS"
由此可见,HADOOP_MAPRED_HOME的值取了2个可能的值:
/opt/cloudera/parcels/CDH/lib/hadoop-mapreduce
或者
/usr/lib/hadoop-mapreduce/
3.查看以上2个目录文件存不存在
# ls /opt/cloudera/parcels/CDH/lib/hadoop-mapreduce
ls: cannot access /opt/cloudera/parcels/CDH/lib/hadoop-mapreduce: No such file or directory
# ls /usr/lib/hadoop-mapreduce/
ls: cannot access /usr/lib/hadoop-mapreduce/: No such file or directory
4.再看看源码
exit 1
+ fi
+
+- if [[ ! -d "${HADOOP_HDFS_HOME}" ]]; then
+- hadoop_error "ERROR: Invalid HADOOP_HDFS_HOME"
+- exit 1
+- fi
+-
+- if [[ ! -d "${HADOOP_YARN_HOME}" ]]; then
+- hadoop_error "ERROR: Invalid HADOOP_YARN_HOME"
+- exit 1
+- fi
+-
+- if [[ ! -d "${HADOOP_MAPRED_HOME}" ]]; then
+- hadoop_error "ERROR: Invalid HADOOP_MAPRED_HOME"
+- exit 1
+- fi
现在确实不存在以上2个目录,所以报了ERROR: Invalid HADOOP_MAPRED_HOME这样的错
5.查找一下hadoop-mapreduce这个目录
/opt/cloudera/parcels/CDH-6.3.2-1.cdh6.3.2.p0.1605554/lib/hadoop-mapreduce
# find / -name hadoop-mapreduce
/opt/cloudera/parcels/CDH-6.3.2-1.cdh6.3.2.p0.1605554/lib/hadoop-mapreduce
6.对比另外一台正常机器的目录
# ls /opt/cloudera/parcels/CDH/lib/hadoop-mapreduce
aliyun-sdk-oss-2.8.3.jar hadoop-mapreduce-client-nativetask-3.0.0-cdh6.3.2.jar
azure-keyvault-core-0.8.0.jar hadoop-mapreduce-client-nativetask.jar
azure-storage-5.4.0.jar hadoop-mapreduce-client-shuffle-3.0.0-cdh6.3.2.jar
bin hadoop-mapreduce-client-shuffle.jar
cloudera hadoop-mapreduce-client-uploader-3.0.0-cdh6.3.2.jar
hadoop-aliyun-3.0.0-cdh6.3.2.jar hadoop-mapreduce-client-uploader.jar
hadoop-aliyun.jar hadoop-mapreduce-examples-3.0.0-cdh6.3.2.jar
hadoop-archive-logs-3.0.0-cdh6.3.2.jar hadoop-mapreduce-examples.jar
另外一台机器有这个目录
7.进到正常机器的/opt/cloudera/parcels查看
# cd /opt/cloudera/parcels
# ls
CDH CDH-6.3.2-1.cdh6.3.2.p0.1605554 LIVY LIVY-0.5.0
明白了,原来是少了个CDH软链接文件,估计是被哪位同事不小心删掉了
看看有问题机器的目录状况:
# cd /opt/cloudera/parcels
# ls
CDH-6.3.2-1.cdh6.3.2.p0.1605554 LIVY LIVY-0.5.0
8.新建软链接解决问题
# cd /opt/cloudera/parcels
# ls
CDH-6.3.2-1.cdh6.3.2.p0.1605554 LIVY LIVY-0.5.0
# ln -s /opt/cloudera/parcels/CDH-6.3.2-1.cdh6.3.2.p0.1605554 /opt/cloudera/parcels/CDH
总结
由上可见具体原因就是少了某个软链接,导致目录不存在,从而导致环境变量取不到值,最终报ERROR: Invalid HADOOP_MAPRED_HOME的错误