Hadoop C访问

引用自：http://blog.csdn.net/johnnywww/article/details/7530364

现在计划做通过C访问Hadoop，选择了LibHDFS，打算用Eclipse(V3.7.2) CDT和cygwin测试。
1.下载CDT，通过Eclipse Install New Software，在地址安装。
2.Cygwin安装开发库，包含gcc, gcc-core,gcc-g++, gcc-mingw-core,  gcc-mingw-g++, make , gdb, binutils
3.添加cygwin的bin目录到Windows的Path环境变量里
4.在Windows下执行gcc，运行报错，显示“访问被拒绝” (access denied)，检查g++.exe, gcc.exe是否只有1k大小，如果是，那么文件是符号链接，那么把文件重命名一下（备份）。然后把g++-3.exe(或者g++-4.exe)拷贝一份，重命名为g++.exe。 把gcc-3.exe拷贝一份，重命名为gcc.exe，参加文章。
5.编写代码

#include "hdfs.h"

int main(int argc, char **argv) {

    if (argc != 4) {
        fprintf(stderr, "Usage: hdfs_write \n");
        exit(-1);
    }

    hdfsFS fs = hdfsConnect("default", 0);
    if (!fs) {
        fprintf(stderr, "Oops! Failed to connect to hdfs!\n");
        exit(-1);
    }

    const char* writeFileName = argv[1];
    tSize fileTotalSize = strtoul(argv[2], NULL, 10);
    tSize bufferSize = strtoul(argv[3], NULL, 10);

    hdfsFile writeFile = hdfsOpenFile(fs, writeFileName, O_WRONLY, bufferSize,
            0, 0);
    if (!writeFile) {
        fprintf(stderr, "Failed to open %s for writing!\n", writeFileName);
        exit(-2);
    }

// data to be written to the file
    char* buffer = malloc(sizeof(char) * bufferSize);
    if (buffer == NULL) {
        return -2;
    }
    int i = 0;
    for (i = 0; i < bufferSize; ++i) {
        buffer[i] = 'a' + (i % 26);
    }
    // write to the file
    tSize nrRemaining;
    for (nrRemaining = fileTotalSize; nrRemaining > 0; nrRemaining -=
            bufferSize) {
        int curSize =
                (bufferSize < nrRemaining) ? bufferSize : (int) nrRemaining;
        hdfsWrite(fs, writeFile, (void*) buffer, curSize);
    }

    free(buffer);
    hdfsCloseFile(fs, writeFile);
    hdfsDisconnect (fs);

    return 0;
}

6.增加环境变量HADOOP_HOME，在cygwin目录下执行

export HADOOP_HOME=/home/test/hadoop0.20.2

或在/etc/profile文件中增加上述内容。

7.在cygwin里面执行

gcc writeHDFS.c -I ${HADOOP_HOME}/src/c++/libhdfs -I /usr/local/jdk/include -I/usr/local/jdk/include/win32 -L${HADOOP_HOME}/c++/Linux-i386-32/lib -lhdfs -o writeHDFS

8.在linux里面执行

gcc writeHDFS.c -I ${HADOOP_HOME}/src/c++/libhdfs -I ${JAVA_HOME}/include -I ${JAVA_HOME}/include/linux  -L${HADOOP_HOME}/c++/Linux-i386-32/lib -lhdfs -L${JAVA_HOME}/jre/lib/i386/client -ljvm -o writeHDFS

9.运行./writeHDFS老报错，提示error while loading shared libraries: libhdfs.so.0: cannot open shared object file: No such file or directory，由于hadoop的c++/Linux-i386-32/lib目录下libhdfs.so、libhdfs.so.0、 libhdfs.so.0.0.0文件相同，先将libhdfs.so.0改为libhdfs.so.0.bak，再执行

ln -s ./libhdfs.so.0.0.0 ./libhdfs.so.0

然后运行

sudo vi /etc/ld.so.conf

添加如下内容

/home/pc01/hadoop-1.0.1/c++/Linux-i386-32/lib
/usr/local/lib/jdk1.7.0_03/jre/lib/i386/server

再运行

sudo /sbin/ldconfig -v