Hoop, Hadoop HDFS over HTTP - Documentation Sets 0.1.0-SNAPSHOT

Hoop has been contributed to Apache Hadoop and it is available as part of Hadoop 0.23.x releases.

For details please refer to the Apache Hadoop site.


Enough reading, take me to the Hoop setup instructions.

Hoop is a server that provides a REST HTTP gateway supporting all HDFS File System operations (read and write).

Hoop can be used to transfer data between clusters running different versions of Hadoop (overcoming RPC versioning issues), for example using Hadoop DistCP.

Hoop can be used to access data in HDFS on a cluster behind of a firewall (the Hoop server acts as a gateway and is the only system that is allowed to cross the firewall into the cluster).

Hoop can be used to access data in HDFS using HTTP utilities (such as curl and wget) and HTTP libraries Perl from other languages than Java.

Hoop also provides a Hadoop HDFS FileSystem implementation to allow that enables access to HDFS over HTTP using the hadoop command line tool as well as the Hadoop FileSystem Java API.

Hoop file URIs are identical to HDFS URIs (except for the SCHEMA://HOST:PORT).

Hoop has built-in security supporting Hadoop pseudo authentication, HTTP SPNEGO Kerberos and pluggable authentication mechanims (authentication is enforced using Alfredo. It also provides Hadoop proxy user support.


Hoop is distributed under Apache License 2.0.

How Does Hoop Works?

Hoop is a separate service from Hadoop NameNode.

Hoop itself is Java web-application and it runs using a preconfigured Tomcat bundled with Hoop binary distribution.

Hoop HTTP web-service API calls are HTTP REST calls that map to a HDFS file system operation. For example, using the curl Unix command:

  • $ curl http://myhoop:14000/user/foo/README.txt returns the contents of the HDFS /user/foo/README.txt file.
  • $ curl http://myhoop:14000/user/foo?op=list returns the contents of the HDFS /user/foo directory in JSON format.
  • $ curl -X POST http://myhoop:14000/user/foo/bar?op=mkdirs creates the HDFS /user/foo.bar directory.

How Hoop and Hadoop HDFS Proxy differ?

Hoop was inspired by Hadoop HDFS proxy.

Hoop can be seening as a full rewrite of Hadoop HDFS proxy.

Hadoop HDFS proxy provides a subset of file system operations (read only), Hoop provides support for all file system operations.

Hoop uses a clean HTTP REST API making its use with HTTP tools more intuitive.

Hoop supports Hadoop pseudo authentication, Kerberos SPENGOS authentication and Hadoop proxy users. Hadoop HDFS proxy does not.

Hadoop HDFS proxy has not been mantained and it may be removed from Hadoop code.

Current Limitations

The FSDataInputStreams returned by the HoopFileSystem client implementation do not support the methods of the Seekable and PositionedReadable interfaces.

The skip() method of InputStreams returned by the HoopFileSystem client skips bytes on the client side, the underlaying inputstream is fully read.