HDFS的API使用

HDFS的API总体来说有两个主要的类。

FileSystem 和Configuration。

导入依赖

            org.apache.hadoop

            hadoop-client

            2.7.3

(1)初始化HDFS

//因为远程提交的情况下如果没有hadoop 的系统环境变量，就会读取当前主机的用户名，所以Hadoop集群的节点中没有该用户名的权限，所以出现的异常。

System.setProperty("HADOOP_USER_NAME","root");

Configuration conf = new Configuration();

        conf.set("fs.defaultFS", "hdfs://master:9000");

        conf.set("dfs.replication", "2");

        FileSystem fs = FileSystem.get(conf);

(2) 上传文件

  fs.copyFromLocalFile(new Path("D:\abc\wordcount\input\test.txt"),new Path("/wordcount/a.txt"));

(3)下载文件

        fs.copyToLocalFile(new Path("/wordcount/a.txt"),new Path("D:\abc\wordcount\input\test2.txt"));

(4)创建

        fs.create(new Path("/wordcount/b.txt"));

(5)删除

        fs.delete(new Path("/wordcount/b.txt"));

(6)获取一个文件的副本数

        short i = fs.getDefaultReplication(new Path("/wordcount/b.txt"));

        System.out.println(i);

(7)获取所有的文件

        RemoteIterator listFiles = fs.listFiles(new Path("/"),true);

        while (listFiles.hasNext()){

            LocatedFileStatus fileStatus = listFiles.next();

            Path path = fileStatus.getPath();

            System.out.println(path);

            //获取Block块的信息

            BlockLocation[] blockLocations = fileStatus.getBlockLocations();

            for (BlockLocation blockLocation:blockLocations) {

                String[] hosts = blockLocation.getHosts();

                String[] names = blockLocation.getNames();

                long offset = blockLocation.getOffset();

                System.out.println(offset);

                System.out.println(ArrayUtils.toString(hosts));

                System.out.println(ArrayUtils.toString(names));

            }

        }

        fs.close();