Cloudera Distribution for Hadoop Review

Cloudera Manager Hadoop Cluster Installation Evaluation

I decided to give Cloudera's Manager software a try, and was pleasantly surprised at how simple it becomes to deploy a substantial Hadoop cluster.

I began by creating an automated kickstart installer for RHEL 6.2 (booting off a custom isolinux image created for this purpose), with all of the required packages, so that from server power on to creating a 20+ node cluster takes less than 15 minutes. The limitation for the number of concurrent node installs is based on network and disk i/o bottlenecks on the deployment server. If you wanted to PXE boot the cluster in a production environment, you would want a bank of servers behind a load balancer, optimally.

Once the Manager is installed on the master node, you simply log into the administration webpage, and from there, add all of the hosts to deploy the cluster on. One nice discovery was that it takes advantage of regular expressions for host names or IP addresses, so you can literally create a cluster containing hundreds of nodes with a trivial amount of effort.

Once the software is deployed, you can select the roles for each of the servers. It's an incredibly painless deployment. That being said, it is not without its flaws.

One of the primary flaws is that all of the configuration and log files are in non-standard locations, and are split in non-standard ways. It's obvious from the way that the files are arranged that it simplifies programmatic deployment. It also makes it a bit harder for a human who is used to standard Hadoop deployments to figure out where everything is located.

And finally, I discovered a bug with one of the packaged software products, Oozie. One of the resource files, oozie-bundle-0.1.xsd contains an invalid regular expression on line 22. I haven't tracked down the behavior, but for some reason JDK 1.6.30 will parse that invalid regex, but JDK 1.7U2 will exit with errors. Naturally, I was running JDK 1.7U2, so it took me a little extra time to debug the problem.

Overall, I quite liked Cloudera's Manager. It's certainly one of the better cluster deployment products I've seen.

**Disclosure: I am a real user, and this review is based on my own experience and opinions.
More Cloudera Distribution for Hadoop reviews from users
...who work at a Financial Services Firm
...who compared it with Oracle NoSQL
Learn what your peers think about Cloudera Distribution for Hadoop. Get advice and tips from experienced pros sharing their opinions. Updated: April 2021.
476,483 professionals have used our research since 2012.
Add a Comment
1 Comment

author avatarit_user217290 (Senior DBA Consultant at a tech services company with 10,001+ employees)


Can I have Cloudera's Manager software for free to test and deploy it on a sandBox to work on a POC purposes.