It federates your distributed storage(HDFS) with the help of a seperate namenode for each federated HDFS partition.
Problem: Say you have 2 tenants for a single cluster hence 2 namenodes, 2 namenodes will imply 2 hive servers, 2 oozie servers and similarly 2 of each of other hadoop components which will communicate with their respective namenode and write in their respective HDFS partition.
compute quotas could be enforced for each tenant.(say tenant 1 - 50% and tenant 2 - 50%)
I could not find how we can develop a similar multi tenant environment for hadoop cluster.
What I was thinking to do was like each tenant(with its users) will have a
All the users belonging to tenant1 group will have access to write in hdfs directory(/usr/tenant1/username)
What problem I faced was.. say i create a table from hive as a tenant1 user, it created the table /apps/hive/warehouse similarly /apps/oozie/data. what I was thinking was it will create table in user's hdfs home directory and hence only tenant1 users will have access to it. but it didn't happen.
Can I get some help in how should I proceed with my development to create multi tenancy over Hadoop ecosystem?