更改hadoop native库文件后datanode故障

hadoop是用cloudra的官方yum源安装的，服务器是CentOS6.3 64位操作系统，自己写的mapreduce执行的时候hadoop会提示以下错误：

WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable

网上说是这样的

Hadoop的native是在32 bit环境下编译的，在64bit环境下运行会有问题，所以需要下载hadoop的源码在64bit环境下重新编译

然后在官方文档里找到了相应的解释

链接：http://hadoop.apache.org/docs/r2.5.1/hadoop-project-dist/hadoop-common/NativeLibraries.html

The pre-built 32-bit i386-Linux native hadoop library is available as part of the hadoop distribution and is located in the lib/native directory. You can download the hadoop distribution from Hadoop Common Releases.

Be sure to install the zlib and/or gzip development packages - whichever compression codecs you want to use with your deployment.

于是，按照官档里的说明在服务器上Build出新的hadoop，把native里面的库文件替换默认的的库文件，搞定，重启hadoop所有服务，自己写的mapreduce可以用了，那么接下来问题来了……发现hdfs进入了Safemode，说有个多少多少块corrupted，然后发现重启期间有同事在使用mapreduce，怀疑是这个原因造成了数据的损坏。由于数据有备份，于是手动解除Safemode，用hadoo fsck删掉了坏掉的数据块，此时是不提示corrupted了，可是惊奇的发现，datanode没有起来！

Hadoop datanode is dead and pid file exists

紧接着，找到存在的pid file，rm之，再起！还是不行！找到datanode的日志，发现有FATAL信息！

2014-11-26 14:35:49,577 FATAL org.apache.hadoop.hdfs.server.datanode.DataNode: Exception in secureMain
java.lang.UnsatisfiedLinkError: org.apache.hadoop.io.nativeio.SharedFileDescriptorFactory.createDescriptor0(Ljava/lang/String;Ljava/lang/St
ring;I)Ljava/io/FileDescriptor;
        at org.apache.hadoop.io.nativeio.SharedFileDescriptorFactory.createDescriptor0(Native Method)
        at org.apache.hadoop.io.nativeio.SharedFileDescriptorFactory.create(SharedFileDescriptorFactory.java:87)
        at org.apache.hadoop.hdfs.server.datanode.ShortCircuitRegistry.<init>(ShortCircuitRegistry.java:169)
        at org.apache.hadoop.hdfs.server.datanode.DataNode.initDataXceiver(DataNode.java:586)
        at org.apache.hadoop.hdfs.server.datanode.DataNode.startDataNode(DataNode.java:773)
        at org.apache.hadoop.hdfs.server.datanode.DataNode.<init>(DataNode.java:292)
        at org.apache.hadoop.hdfs.server.datanode.DataNode.makeInstance(DataNode.java:1893)
        at org.apache.hadoop.hdfs.server.datanode.DataNode.instantiateDataNode(DataNode.java:1780)
        at org.apache.hadoop.hdfs.server.datanode.DataNode.createDataNode(DataNode.java:1827)
        at org.apache.hadoop.hdfs.server.datanode.DataNode.secureMain(DataNode.java:2003)
        at org.apache.hadoop.hdfs.server.datanode.DataNode.main(DataNode.java:2027)
2014-11-26 14:35:49,580 INFO org.apache.hadoop.util.ExitUtil: Exiting with status 1
2014-11-26 14:35:49,582 INFO org.apache.hadoop.hdfs.server.datanode.DataNode: SHUTDOWN_MSG:
/************************************************************
SHUTDOWN_MSG: Shutting down DataNode at carweb94/10.14.1.94

搜索了下，发现是native库文件的问题！好吧，这下找到元凶了，把默认的native库文件替换回来了，问题解决了。排错的过程中，这篇文章给了很大启发。

http://stackoverflow.com/questions/26467568/hadoop-2-5-0-failed-to-start-datanode