Are Apache HBase and Cloudera HBase compatible?

[Brian Zambrano <br...@eventbrite.com>]

At work we are attempting to do the following:

- Run Elastic MapReduce jobs via Amazon, which freezes Hadoop at version
0.20.205
- Write output to HBase running on EC2, specifically, 0.92.1-cdh4.0.1 from
Cloudera

What I've discovered so far is my WordCount test appears to work when I
package Apache HBase 0.92.1 in my Hadoop job (via maven). I'm scared that
this is working accidentally and that may blow up as my HBase usage matures
(it's just doing basic Puts now).

However, when I package HBase 0.92.1-cdh4.0.1 in my Hadoop job, I get a
ClassNotFoundException:

https://emr-qa.eventbrite.com.s3.amazonaws.com/logs/j-RWJ75VR11SLB/steps/1/stderr

So, it appears that Hadoop 0.20.205 and HBase 0.92.1-cdh4.0.1 don't like
each other.  Since Hadoop 0.20.205 and Apache HBase 0.92.1 appear to work
together:

- Does the Apache HBase jar play nicely with the CDH Hbase server?
- Is mixing versions/distributions like this is horrible idea?

BZ

Re: Are Apache HBase and Cloudera HBase compatible?

[Jean-Daniel Cryans <jd...@apache.org>]

Using a HBase version compiled on Hadoop 2.0 (like CDH4) and trying to
use it on Hadoop 1.0 or previous is bound to fail. You could try
recompiling CDH4's HBase on 0.20.205 but I don't know if it will work,
FWIW Apache's HBase is compiled on Hadoop 1.0 which is "close" to
0.20.205 and it "worked" for you.

So:

> - Does the Apache HBase jar play nicely with the CDH Hbase server?

If they are compiled on the same Hadoo major version I'd say that in
general it should work.

> - Is mixing versions/distributions like this is horrible idea?

Mixing those Hadoop major versions is, but the HBase versions you are
using aren't really that different.

Hope this helps,

J-D