Monday, June 17, 2013

Setting up Secure Cluster - Evaluation of Products

I am trying to setup a secure Hadoop cluster. I chose the stable Hadoop version 1.1. I have setup Hadoop on my 6 node cluster.

Now my primary goal is to bring up a secure cluster and then we can request more machines and rollout our cluster. I am basically doing a POC with my 6 node cluster. I plan to support only HDFS and MapReduce.

Per Hadoop documentation, to enable a secure cluster , I need Kerberos and an LDAP server. Kerberos is for Authentication and LDAP is for Group Information. TO enable LinuxController for running mapreduce tasks , I need to have LDAP on cluster nodes as well.

I am familiar with LDAP , but not with Kerberos.  Now I need to installa MIT Keberos.

Before setting up MIT Kerberos, I want to check if anything else is avalable. I get this product - whosso  @ http://whosso.com/

The product enables Hadoop Security without Kerbros and LDAP. I am happy. Thanks Google (again) .

I tried the product by following thier installation manual and setting up Hadoop security by entering user information and group information in files. Worked right out of the box.

But I like to hookup LDAP (Active Directory) to Whosso product so I can use the user information already stored in Active Directory. Looks like Whosso allows me to plugin my AuthenticationProvider .I haven't tried it yet.
For Hadoop servers, I plan to keep the information in files.

I also like it  if whosso could setup the use rinformation in database instead of files as a standard plugin.  will post in their forum.

ANyway , pretty happy with my secure cluster. No LDAP, No Kerberos