Install Hue without Cloudera

Question

Has anyone tried/succeeded in installing Hue on Hadoop without Cloudera?

I have gotten to a point where I can reliably reproduce a hadoop cluster with hbase and hive and can set it all up in about 15 minutes. I'd love to have Hue along with all this without having to go back and redo my setup with Cloudera.

Answer 1

Follow the Hue manual installation steps from Hortonworks documentation, it will take you step-by-step on how to do it manually.

Answer 2

Checkout slides #19 & #5, Hue is getting everywhere and is compatible with Hadoop 0.20 / 1.2.0 / 2.2.0: http://gethue.com/hue-goes-to-paris-hug-france/

Hue has tarball releases releases that you are free to install. You can also simply clone the source code (Hue is open source and Apache Licenced) github: https://github.com/cloudera/hue and build the branch you want.

Upstream documentation is here or CDH's one here.

Hue is also packaged in BigTop (and so based on Vanilla Hadoop).

Hue is a Web Server (Django based) which acts as a view on top of Hadoop. So Hue just needs to be installed and then configured by adding the hosts of NameNode, JobTracker, Resource Manager, Oozie, HiveServer... etc in its hue.ini.

Also, as detailed on the gehue.com/releases, the version you need might depend on your Hive version.

Notice that without Cloudera's distribution your mileage might vary but feel free to chime-in on the Hue user-list or gethue.com ;)

We are also seeing for improving Hue setup with Amazon AWS/EMR!

Answer 3

To build and run hue 3.6.0 with apache hadoop 2.4.1

1. git clone https://github.com/cloudera/hue.git (Notice! releases/tag/release-3.6.0 is unstable, It's better to build from latest master. I built from Aug 7, 87d6b2da1 - it's stable)

2. cd hue

3. $ vi maven/pom.xml

4. change hadoop.version to 2.4.1
5. replace hadoop-core with hadoop-common

6. set hadoop-test version to 1.2.1

7. remove files which need hadoop mr1

8. $ rm desktop/libs/hadoop/java/src/main/java/org/apache/hadoop/mapred/ThriftJobTrackerPlugin.java

9. $ rm desktop/libs/hadoop/java/src/main/java/org/apache/hadoop/thriftfs/ThriftJobTrackerPlugin.java

10. build hue $ make apps

11. configure hue $ vi desktop/conf/pseudo-distributed.ini

12. run hue server in dev mode $ build/env/bin/hue runserver 0.0.0.0:8000

Answer 4

Quote: "...without Cloudera's distribution your mileage might vary...."

Indeed, it will vary A LOT! It would seem that the following is quite true:

Per the install giude: http://cloudera.github.io/hue/docs-2.0.1/manual.html#_install_hue

NOTE: Hue requires the Hadoop contained in Cloudera’s Distribution including Apache Hadoop (CDH), version 3 update 4 or later.

I've tried it and have run into walls with Hue trying to connect to Hive, Pig and OOZIE.

At this stage - from my experience at least - Hue will NOT run on a standard Apache Hadoop installation using standard Apache tools like Hive and Pig. It must be a vintage of Cloudera’s Distribution.

If anyone has any other (positive) experiences installing Hue outside of the Cloudera’s Distribution, I'd be quite interested to hear about them...