Mini Hadoop Cluster Notes

This is a Work In Progress page...

The goal is to build a 4-node HDP cluster on some Intel NUC machines.  There will be a single master node built on a NUC7i5BNH and three worker nodes using older/cheaper NUCs as described in unboxing my new little box (my first intel nuc).

First, I installed Ubuntu 16.04 with help from https://tutorials.ubuntu.com/tutorial/tutorial-create-a-usb-stick-on-macos.

Then I set a root password and allowed login screen access; http://www.configserverfirewall.com/ubuntu-linux/enable-ubuntu-desktop-root-login/ and https://askubuntu.com/questions/951581/how-to-enable-ssh-root-access-ubuntu-16-04.

Enabled ssh access for root; http://ubuntuhandbook.org/index.php/2016/04/enable-ssh-ubuntu-16-04-lts/.

Allow 22 and 5900 (SSH and VNC) to be punch through to master1 (192.168.0.151) after setting up reservations described at http://uk.tp-link.com/faq-182.html for the following bits (need their MAC addresses to do this).

  • master1 - 192.168.0.151
  • worker1 - 192.168.0.161
  • worker2 - 192.168.0.162
  • worker3 - 192.168.0.163

Enable VNC - http://ubuntuhandbook.org/index.php/2016/07/remote-access-ubuntu-16-04/.


OK... OK... HERE IS THE BASIS OF MY PENDING BLOG POST ABOUT GETTING THE NETWORK ROLLING!!

Blog Posting Title

building a mac-hosted network for a nuc cluster (leveraging a tp-link n300 nano router)

Blog Posting Itself



As I called out at the bottom of unboxing my new little box (my first intel nuc), my real goal was to create a Hadoop cluster based on Intel NUC machines.  My thinking has always been to be able to have a small travel case for a few NUCs to take with me on the road which is the reason that blog post also mentioned I needed to figure out "some creative (and portable) networking" and this post is to document what I ended up doing.  First up, let's talk hardware!

NUC Specifications

I liked the price point of the NUC5CPYH so much I decided to buy two more identical setups.  The price for each of these was a bit higher at $245 instead of the original $215 I spent.  These eventually are going to be my worker nodes in my Hadoop cluster.  I decided I needed something much more beefy for my master node, so I went with this setup.

That came in just under $780, but I actually burned some funny-money in the form of Amazon points I've been racking up from all of my traveling.  That puts the price of hardware for this 4-node cluster of NUC machines just under $1500.

Installing Ubuntu

TALK ABOUT SETTING UP 16.04 AND NAMING MACHINES master1.nuc and worker1.nuc - worker3.nuc, CONNECTING TO MY HOME WIRELESS NETWORK.  ENABLING ROOT ACCESS, AND SSH FOR ALL 3 BOXES, BUT VNC ONLY FOR MASTER1.

The Network

As the title suggests, my goal was to have a network that would be reliant on my MacBook Pro 15 that I travel with (ok... any Mac for that matter as I'm thinking I'll hook it all up to an old MacBook Air when it stays at home).  TELL ABOUT HOW TO SET THIS UP.

http://uk.tp-link.com/download/TL-WR802N_V2.html

THEN TELL ABOUT TP-LINK ROUTER IN DEFAULT MODE.  GETTING THE BETTER NUC ON THIS WIFI NETWORK, LOCKING IN 192.168.0.151, PUNCHING DOWN THE PORTS AND CONNECTING VIA SSH, VNC, PLUS OPENING UP ADDITIONAL PORTS I'LL NEED LATER.  /ETC/HOSTS IN MAC FOR 192.168.2.2 AND VALIDATE CONNECTIVITY WORKS.

DON'T FORGET NEED TO FOLLOW INSTRUCTIONS AT https://askubuntu.com/questions/894750/allow-a-guest-session-wifi-access-with-no-users-logged-in TO ALLOW LOGIN TO WIFI B4 USER LOGS IN.

Pulling It All Together

ON MASTER1, ADD LOCAL /ETC/HOSTS ENTRIES FOR ITSELF AND ADDITIONAL 3 WORKERS.  VALIDATE FORWARD & REVERSE DNSCONFIGURE 3 WORKERS (ONE AT A TIME) WITH ADDR RESERVATIONS.  POINT OUT NO NEED FOR VNC ON THESE MACHINES.  JUST USING SSH AND GO AHEAD AND SETUP PASSWORD-LESS SSH ROOT ACCESS WITH LINK TO HDP PAGE FOR DETAILS.

POINT OUT ALL IS ROLLING NOW AS FAR AS OS & NETWORK!!

Next Steps

Well... it is time to install Hadoop on this mini NUC-based cluster which I'll capture in a follow-on blog posting (I'll add a comment below pointing to it once it is published).  Also... I'm thinking I'll start looking for a rugged carry case with plenty of foam so I can take this cluster on the road with me.  This one looks pretty decent.