Hadoop web编程--REST API

1 介绍

    Hadoop提供了一个Java native API来支持对文件系统进行创建，重命名，删除文件或者目录，打开读取或者写文件，设置文件权限等操作。这对于运行在hadoop集群中的应用程序来说是挺棒的，但是，也有许多外部的应用程序需要操作HDFS的情况，怎么办？如果解决这种问题呢？Hortonworks 开发了一些额外的API来支持这些基于标准REST功能的需求。

2 WebHDFS REST API

    WebHDFS观念是基于HTTP操作，比如GET、PUT、POST和DELETE。像OPEN、GETFILESTATUS、LISTSTATUS的操作是使用HTTP GET，其它的像CREAT、MKDIRS、RENAME、SETPERMISSION是依赖于HTTP PUT类型。APPEND操作时基于HTTP POST类型，然而，DELETE是使用HTTP DELETE。

    认证方式可以使用基于user.name参数或者如果安全机制开启了就依赖于Kerberos。标准的URL格式如下所示：

 http://host:port/webhdfs/v1/?op=operation&user.name=username

默认的启动对口是14000，你可以在httpfs-env.sh 中配置端口的值。所有与httpfs有关的环境参数变量，你可以再httpfs-env.sh中进行个性化的配置。

3 Httpfs 配置启动

    编辑core-site.xml文件，添加如下内容：

…

  <property>

    <name>hadoop.proxyuser.#HTTPFSUSER#.hosts</name>

    <value>*</value>

  </property>

  <property>

    <name>hadoop.proxyuser.#HTTPFSUSER#.groups</name>

    <value>*</value>

  </property>

….

值得注意的是，#HTTPFSUSER#指的是用户名，即linux启动httpfs的用户。

编辑hdfs-site.xml文件，添加下列属性配置。

<property>

          <name>dfs.webhdfs.enabled</name>

           <value>true</value>

</property>

<property>
<name>dfs.permissions</name>
<value>false</value>
</property>

    另外，想知道更多的有关HDFS端口信息，请参考Cloudera blog.

    启动httpfs，执行如下命令：

  

 httpfs.sh start

4 验证

一切配置好了之后，测试一下是否配置成功是非常有必要的。有这样一种情况，我们需要知道hdfs 目录下tmp的文件状态。按照这样的需求，我们可以编写这样的命令：

命令行下

 curl –i http://hadoop-master:14000/webhdfs/v1/tmp?user.name=app&op=GETFILESTATUS
 HTTP/1.1 200 OK
 Server:Apache-Coyote/1.1
 Set-Cookie:hadoop.auth="u=app&p=app&t=simple&e=1393293979907&s=9YBXB7GM3oiIOYgflAjlq3GIpxc=";Version=1; Path=/
 Content-Type:application/json
 Transfer-Encoding:chunked
 Date: Mon, 24Feb 2014 16:06:20 GMT

 {"FileStatus":{"pathSuffix":"","type":"DIRECTORY","length":0,"owner":"app","group":"supergroup","permission":"720","accessTime":0,"modificationTime":1391352186043,"blockSize":0,"replication":0}}

浏览器中：

 http://hadoop-master:14000/webhdfs/v1/tmp?user.name=app&op=GETFILESTATUS

返回信息：

  
  
  
  
{"FileStatus":{"pathSuffix":"","type":"DIRECTORY","length":0,"owner":"app","group":"supergroup","permission":"720","accessTime":0,"modificationTime":1391352186043,"blockSize":0,"replication":0}}
下面主要介绍几种常用的WebHDFS操作。
5.1 创建
     创建一个目录/tmp/webhdfs。

  
  
  
  
<span style="font-size:18px;">curl-i -X PUT"http://hadoop-master:14000/webhdfs/v1/tmp/webhdfs?user.name=app&op=MKDIRS"
HTTP/1.1 200 OK
Server: Apache-Coyote/1.1
Set-Cookie:hadoop.auth="u=app&p=app&t=simple&e=1393356198808&s=74NhnIdH7WceKqgTW7UJ1ia9h10=";Version=1; Path=/
Content-Type: application/json
Transfer-Encoding: chunked
Date: Tue, 25 Feb 2014 09:23:23 GMT</span>


达到相似的功能，Hadoop 命令：

  
  
  
  
<span style="font-size:18px;">hdfs dfs -mkdir /tmp/webhdfs</span>


创建一个文件
 
创建一个文件需要两个步骤：第一步是在namenode运行命令，第二步根据第一步提供的location参数执行PUT操作。

  
  
  
  
<span style="font-size:18px;">curl -i -X PUT"http://hadoop-master:50070/webhdfs/v1/tmp/webhdfs/webhdfs-test.txt?user.name=app&op=CREATE"
HTTP/1.1 307 TEMPORARY_REDIRECT
Cache-Control: no-cache
Expires: Thu, 01-Jan-1970 00:00:00 GMT
Date: Wed, 26 Feb 2014 14:35:29 GMT
Pragma: no-cache
Date: Wed, 26 Feb 2014 14:35:29 GMT
Pragma: no-cache
Set-Cookie:hadoop.auth="u=app&p=app&t=simple&e=1393461329238&s=as2cO3rRtvj8psr0jAhMk6YBHRY=";Path=/
Location:http://machine-2:50075/webhdfs/v1/tmp/webhdfs/webhdfs-test.txt?op=CREATE&user.name=app&namenoderpcaddress=hadoop-master:9000&overwrite=false
Content-Type: application/octet-stream
Content-Length: 0
Server: Jetty(6.1.26)</span>


发送数据到指定文件中：

  
  
  
  
<span style="font-size:18px;">curl -i -X PUT -T webhdfs-test.txt"http://machine-2:50075/webhdfs/v1/tmp/webhdfs/webhdfs-test.txt?op=CREATE&user.name=app&namenoderpcaddress=hadoop-master:9000&overwrite=false"
HTTP/1.1 100 Continue

HTTP/1.1 201 Created
Cache-Control: no-cache
Expires: Wed, 26 Feb 2014 14:37:51 GMT
Date: Wed, 26 Feb 2014 14:37:51 GMT
Pragma: no-cache
Expires: Wed, 26 Feb 2014 14:37:51 GMT
Date: Wed, 26 Feb 2014 14:37:51 GMT
Pragma: no-cache
Location:webhdfs://0.0.0.0:50070/tmp/webhdfs/webhdfs-test.txt
Content-Type: application/octet-stream
Content-Length: 0
Server: Jetty(6.1.26)</span>


5.2 读取文件
    读取/input文件信息，使用-L参数。

  
  
  
  
<span style="font-size:18px;">[app@hadoop-master ~]$ curl -i -L"http://hadoop-master:14000/webhdfs/v1/input?op=OPEN&user.name=app"
HTTP/1.1 200 OK
Server: Apache-Coyote/1.1
Set-Cookie:hadoop.auth="u=app&p=app&t=simple&e=1393420279602&s=dSnnx9oOwgGwVV/Q/ZmFyRjbtFU=";Version=1; Path=/
Content-Type: application/octet-stream
Content-Length: 1358
Date: Wed, 26 Feb 2014 03:11:19 GMT
Apache HBase [1] is an open-source,distributed, versioned, column-oriented
store modeled after Google' Bigtable: ADistributed Storage System for
Structured Data by Chang et al.[2]  Just as Bigtable leverages the distributed
data storage provided by the Google FileSystem, HBase provides Bigtable-like
capabilities on top of Apache Hadoop[3].

To get started using HBase, the fulldocumentation for this release can be
found under the doc/ directory thataccompanies this README.  Using abrowser,
open the docs/index.html to view theproject home page (or browse to [1]).
The hbase 'book' at docs/book.html has a'quick start' section and is where you
should being your exploration of thehbase project.

The latest HBase can be downloaded froman Apache Mirror [4].

The source code can be found at [5]

The HBase issue tracker is at [6]

Apache HBase is made available under theApache License, version 2.0 [7]

The HBase mailing lists and archives arelisted here [8].

1. http://hbase.apache.org
2.http://labs.google.com/papers/bigtable.html
3. http://hadoop.apache.org
4.http://www.apache.org/dyn/closer.cgi/hbase/
5.http://hbase.apache.org/docs/current/source-repository.html
6.http://hbase.apache.org/docs/current/issue-tracking.html
7.http://hbase.apache.org/docs/current/license.html
8. http://hbase.apache.org/docs/current/mail-lists.html</span>


 
5.3 重命名目录
   需要修改op的值和添加destination参数，实例如下所示：

  
  
  
  
<span style="font-size:18px;">[app@hadoop-master~]$ curl -i -X PUT
"http://hadoop-master:14000/webhdfs/v1/tmp/webhdfs?op=RENAME&user.name=app&destination=/tmp/webhdfs-new"
[app@hadoop-master~]$ curl -i -X PUT
"http://hadoop-master:14000/webhdfs/v1/tmp/webhdfs?op=RENAME&user.name=app&destination=/tmp/webhdfs-new"
HTTP/1.1 200 OK
Server:Apache-Coyote/1.1
Set-Cookie: hadoop.auth="u=app&p=app&t=simple&e=1393420506278&s=Oz9HEzxuYvP8kfAY4SWH6h+Gb50=";Version=1; Path=/
Content-Type:application/json
Transfer-Encoding:chunked
Date: Wed, 26 Feb2014 03:15:06 GMT

{"boolean":true}</span>


验证是否正确地执行，输入下列命令：

  
  
  
  
<span style="font-size:18px;">[app@hadoop-master~]$ hdfs dfs -ls /tmp
Found 5 items
drwx------   - app supergroup          0 2014-01-08 08:02 /tmp/hadoop-yarn
drwxr-xr-x   - app supergroup          0 2014-02-02 09:43 /tmp/hdfs_out
drwxr-xr-x   - app supergroup          0 2014-02-20 22:39 /tmp/hive-app
drwxr-xr-x   - app supergroup          0 2014-02-25 04:25 /tmp/jps
drwxr-xr-x   - app supergroup          0 2014-02-25 04:23 /tmp/webhdfs-new</span>


5.4 删除目录
非空的目录删除的话会抛出异常，只有为空的目录才会被删除。

  
  
  
  
<span style="font-size:18px;">[app@hadoop-master ~]$ curl -i -X DELETE
"http://hadoop-master:14000/webhdfs/v1/tmp/webhdfs-new?op=DELETE&user.name=app"
HTTP/1.1 200 OK
Server: Apache-Coyote/1.1
Set-Cookie:hadoop.auth="u=app&p=app&t=simple&e=1393421161052&s=r03LOm2hO91ujcc66wNWyMJnDx4=";Version=1; Path=/
Content-Type: application/json
Transfer-Encoding: chunked
Date: Wed, 26 Feb 2014 03:26:01 GMT

{"boolean":true}</span>


核查是否正确地执行

  
  
  
  
<span style="font-size:18px;">[app@hadoop-master ~]$ hdfs dfs -ls /tmp
Found 4 items
drwx------   -app supergroup          0 2014-01-0808:02 /tmp/hadoop-yarn
drwxr-xr-x   -app supergroup          0 2014-02-0209:43 /tmp/hdfs_out
drwxr-xr-x   -app supergroup          0 2014-02-2022:39 /tmp/hive-app
drwxr-xr-x   -app supergroup          0 2014-02-2504:25 /tmp/jps</span>


总结
   WebDFS提供了一个简单、标准的方式来执行Hadoop 文件系统操作，这个客户端不必运行在Hadoop集群本身中。WebHDFS最主要的特性是让客户端通过预定义端口直接地链接namenode 和 datanode。这样的话，规避了好多HDFS proxy的代理层以及预先配置Tomcat的绑定。WebHDFD API是可互相交换的，这样让客户端不需要去打开防火墙端口。
 
两种常见的异常:
1. HTTP/1.1 405 HTTP method PUT is not supported by this URL
   修改hdfs-site.xml文件的权限属性。

  
  
  
  
<property>
<name>dfs.permissions</name>
<value>false</value>
</property>

2 {"RemoteException":{"exception":"AccessControlException","javaClassName":"org.apache.hadoop.security.AccessControlException","message":"Permission denied: user=dr.who, access=WRITE, inode=\"/qzhang\":Administrator:supergroup:drwxr-xr-x\n\tat
   问题2与问题1是一个问题，解决方案同上。