在《Hadoop三:Hadoop Java API初探(完结)》中已经介绍了为什么需要从winddows的eclipse转到linux下eclipse+maven来写java api,以及在linux搭建开发环境的demo程序。(详见:https://www.jianshu.com/p/dd13c1dba52d)
这一节我会介绍更详细的HDFS客户端API程序
写代码的时候需要从官网查看api文档,地址:https://hadoop.apache.org/docs/r2.9.1/api/
1. 初始化部分代码
conf.set("fs.defaultFS", "hdfs://10.10.77.194:9000"); //这一行的作用很重要,因为new出来的conf文件仅仅存储的是hdfs配置文件默认值,必须要在客户端手动设置
conf.set("dfs.replication", "2"); //另外需要根据实际情况设置每一个block的副本数量,我做实验设置的是2(配置文件中也是设置的2,跟配置文件保持同步就好)
package com.gamebear.test1;
import java.net.URI;
import java.util.Iterator;
import java.util.Map.Entry;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.FileStatus;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.LocatedFileStatus;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.fs.RemoteIterator;
public class HadoopHdfs {
static FileSystem fsObj = null;
static Configuration conf = null;
private static void init() throws Exception {
conf = new Configuration(); /* set the "http://hadoop.apache.org/docs/stable/api/index.html" * set core-site.xml, "10.10.77.194" is the nameNode, "9000" is the nameNode client port */
conf.set("fs.defaultFS", "hdfs://10.10.77.194:9000");
conf.set("dfs.replication", "2");
//conf.set("fs.hdfs.impl", "org.apache.hadoop.hdfs.DistributedFileSystem");
System.out.println(conf.get("fs.hdfs.impl")); fsObj = FileSystem.get(conf); //get a hadoop FileSystem client object, refer to the web:"http://hadoop.apache.org/docs/stable/api/index.html"
//fsObj = FileSystem.get(new URI("hdfs://10.10.77.194:9000"), conf, "root"); //get the configuration file from hadoop server
}
2. 查看默认配置参数
需要注意的是,有些参数是Hadoop服务器端参数,有些是客户端参数
private static void testConf() throws Exception{
//import java.util.Map.Entry;
Iterator<Entry<String, String>> it = conf.iterator();
while(it.hasNext()){
Entry<String, String> ent = it.next();
System.out.println(ent.getKey() + ":" + ent.getValue());
}
}
关注如下默认参数:
dfs.blocksize:134217728 //默认值128M,最小值是1M
dfs.replication:3 //默认值3, 我这里测试只有2台datanode,设置为了2
dfs.datanode.data.dir:file://${hadoop.tmp.dir}/dfs/data
3. 上传下载文件
在linux环境下的eclipse没有跨平台的问题
private static void testUpload() throws Exception
{
fsObj.copyFromLocalFile(new Path("/kluter/hdpTestFile"), new Path("/json201710.tgz.copy"));
System.out.println("upload finished!");
fsObj.close();
}
private static void testDownload() throws Exception
{
fsObj.copyToLocalFile(new Path("/json201710.tgz.copy"), new Path("/kluter/hdpTestFileDl"));
System.out.println("download finished!");
fsObj.close();
}
4. 嵌套创建目录结构
private static void testMkDir() throws Exception
{
boolean mkdirs = fsObj.mkdirs(new Path("/aaa/bbb/ccc"));
System.out.println("mkdirs finished!");
}
5. 递归删除目录
private static void testDel() throws Exception
{
boolean delBl = fsObj.delete(new Path("/aaa/bbb"), true);
System.out.println(delBl);
}
6. 递归查看文件夹下的文件信息
private static void testLs() throws Exception{
RemoteIterator fileLst = fsObj.listFiles(new Path("/"), true);
while(fileLst.hasNext()){
LocatedFileStatus fileStatus = fileLst.next();
System.out.println("blocksize: " + fileStatus.getBlockSize());
System.out.println("owner: " + fileStatus.getOwner());
System.out.println("Replication: " + fileStatus.getReplication());
System.out.println("permission: " + fileStatus.getPermission());
System.out.println("permission2: " + fileStatus.getPath().getName());
BlockLocation[] blockLocations = fileStatus.getBlockLocations();
for(BlockLocation bl:blockLocations){
System.out.println("block-len: " + bl.getLength() + "---" + "block-offset:" + bl.getOffset());
String[] hosts = bl.getHosts();
for(String host:hosts){
System.out.println(host);
}
}
}
}
这里使用Iterator而不使用List的原因是因为,如果hdfs中有上亿个文件,那么客户端使用List的话内存容易爆掉,而且网络传输也很久。
而迭代器只是一种取数据的方式,调用next的时候才分配一个文件的少量内存,不占用大内存空间
代码打印出了分block文件的信息,查看block所在的host,便于mapreduce的时候分布式计算
如果不想递归查看,参数写false
7. 查看单目录下文件信息
private static void testLs2() throws Exception {
FileStatus[] listStatus = fsObj.listStatus(new Path("/"));
for(FileStatus file:listStatus) {
System.out.println("permission2: " + file.getPath().getName());
System.out.println("permission2: " + file.getPath());
System.out.println(file.isDir()?"directory":"file");
System.out.println("----------------------------------------------");
}
}
这种方式适用于单个目录下文件较少的情况
8. 使用stream方式上传、下载文件
用流的方式来操作hdfs上的文件,可以实现读取指定偏移量范围的数据
8.1 使用stream方式上传文件
将/kluter/hdpTestFile以steam的形式上传到hdfs的aaa目录下:
private static void testUpload() throws Exception{
FSDataOutputStream dfsOutStream = fsObj.create(new Path("/streamTestFile"), true);
FileInputStream inputStream = new FileInputStream("/kluter/streamTestFile");
IOUtils.copy(inputStream, dfsOutStream);
}
8.2 使用stream方式下载文件
private static void testDownload() throws Exception{
FSDataInputStream inputStream = fsObj.open(new Path("/streamTestFile"));
FileOutputStream dfsOutStream = new FileOutputStream("/kluter/downloadStreamFile");
IOUtils.copy(inputStream, dfsOutStream);
}
9. 使用stream方式,按需求读取
private static void testRamdomAccess() throws Exception{
FSDataInputStream inputStream = fsObj.open(new Path("/streamTestFile"));
FileOutputStream OutStream = new FileOutputStream("/kluter/downloadStreamFile");
/**
* just seek 100, pass the first 100 bytes
*/
// inputStream.seek(100);
// IOUtils.copy(inputStream, OutStream);
/**
* copy by client control, u can use while here
*/
IOUtils.copyLarge(inputStream, OutStream, 100, 300);
}
private static void testPrintScr() throws Exception{
FSDataInputStream inputStream = fsObj.open(new Path("/streamTestFile"));
IOUtils.copy(inputStream, System.out);
}