Clustered Pumpkins

This is the second step in our journey to do cool things by combining Pi with Akka/Scala, bringing us closer to making fully interactive pumpkins! This time we get Akka Cluster going with 2 raspberry pi's and my laptop as a 3rd node just to show off the versatility of Akka/Scala. Check out this sweet gif of two pi's and a laptop in a clustered setup (The Raspberry pi's are at the top, my laptop is the bottom pane):

What does this do?

It uses the power of Akka Clustering to allow communication amongst many Raspberry Pi computers, complete with built-in gossipping, cluster state convergance, and fault-tolerance (it gracefully handles member nodes joining and leaving at any time). This example also has some custom messaging built in. Every node in the cluster periodically sends runtime information about itself to another node in the cluster (using Akka routing).

This demo shows a node (my laptop) being forcefully taken out of service. Notice how both the Raspberry Pi nodes instantly respond by marking the failed node as temporarily down and eventually fully terminated (Pi's are the top 2 panes, laptop is the bottom):

Running the example

Building a fat jar

The steps to clone/build/assemble the project are identical to the steps outlined in the hello-akka application so head over there first if you need a refresher on that.

Starting the cluster

Once you have the fat jar distributed everywhere (ideally you want 2-3 rasberry pi's + maybe a laptop running this) we can start a new node with this command:

java -jar clustered-pi-assembly-1.0.jar \
  192.168.1.156:2551 \ # First seed node
  192.168.1.198:2551 \ # Second seed node
  192.168.1.198 \ # The IP of the node you are currently SSH'd to
  2551 # The port that you want the node to run on

Let's break this down. In Akka clustering, you need at least 2 seed nodes to start the cluster. You tell the node about them by passing them as command line arguments in the for of hostname:port (can be an IP address too). The 3rd and 4th arguments is the external hostname/IP address and Port for the node you are currently on. The order you start the nodes doesn't really matter that much since the cluster will self-heal but here's an example:

First I'll login to a Pi with IP 192.168.1.198 and run this (I'll make this guy a seed node):

java -jar clustered-pi-assembly-1.0.jar \
  192.168.1.156:2551 \ # First seed node (IP of my laptop)
  192.168.1.198:2551 \ # Second seed node (IP of this node)
  192.168.1.198 \
  2551

Next, lets start up the second seed node on my laptop (notice how the seed nodes are always the same):

java -jar clustered-pi-assembly-1.0.jar \
  192.168.1.156:2551 \ # First seed node (IP of my laptop)
  192.168.1.198:2551 \ # Second seed node (IP of initial Pi node)
  192.168.1.156 \
  2551

Lastly let's start a node on my 2nd pi (I only have two at them moment):

java -jar clustered-pi-assembly-1.0.jar \
  192.168.1.156:2551 \ # First seed node (IP of my laptop)
  192.168.1.198:2551 \ # Second seed node (IP of initial Pi node)
  192.168.1.200 \
  2551

There you go! If all goes well then you should see cluster log output for all 3 nodes. Make sure to play around by adding/removing nodes and watch the cluster handle the failures and changes in cluster state.