Reactive Missile Defend

Reactive Missile Defend is an application to demonstrate how Akka Cluster (Sharding and Singleton) works in stressful situation like network or power problem. The application is simulation of missile defence system. To intercept enemy missiles/bombs we are using missile towers. Every tower is seperate instance of Akka actor. These actors are distributed on many nodes. You can kill one of nodes and watch how actors are migrated.

Game looks like this:

Presentation

Recording from talk is available Scala by the bay youtube channel

Defence

Our defence missile towers (one actor per each) and command centers (cluster nodes on which tower actors are run).

Towers can be in state:

Ready to fire missile

Realoding after firing missile

Offline - UI does not receive message from this tower for more than 500ms

Infected by enemy virus, cant fire missile.

Command centers cluster nodes with started Sharding for TowerActor. Command centers are displayed in form:

IP => number of tower actors running

In case of running on local host:

PORT => number of tower actors running

Under the tower is an icon. This icon represent one of command centers. It visualize on which node tower actor is running. Move mouse cursor on command center and towers will be highlighted.

And can be in color:

• Green - node is up and running
• Orange - Akka cluster detects that node is unreachable.
• Red - Akka cluster removes node from cluster or node has left cluster

Alien weapons

Alien missile - fast, small radius and low damage

Bomb - slow small radius and medium damage

Nuke - big damage and radius

Virus - Infect defence towers, move actor to Infected state for a few second. In this state tower can't shoot (actor is ingoring messages)

Slides from Krakow Scala User Group

How cluster looks like

Cluster should have following nodes:

• Swing UI with Game engine (sbt ui)
• Command centers to run defence tower logic (choosing target and calculating interception vector) (sbt cc)
• Optional CLI UI (sbt cliui)

Requirements

Cassandra as persistence

This project uses Cassandra 2 or 3 as a persistence. You can run cassandra using Docker or connect to existing Cassandra cluster. Details of Cassandra docker image can be found here. Two keyspaces will be created:

• rmd_journal - for journal
• rmd_snapshot - for snapshots These keyspaces can be deleted between running game.

Using Cassandra from docker

docker pull cassandra:3.1

docker run -d -p 9042:9042 --name cass cassandra:3.1

If you are Windows or OSX user you have to add port forwarding for port 9042.

Specify Cassandra host using environment variables

By default Cassandra host is 127.0.0.1

export=CASSANDRA_HOST=myIp

How to run only local nodes (using SBT):

First start Cassandra which will be used as a persitence for Actors. You can run Cassadnra using docker or downloaded binaries.

Run Swing UI. This node is also game engine. Do not kill it. In terminal run

• export SEED_NODE=akka.tcp://[email protected]:3000
• export SEED_NODE2=akka.tcp://[email protected]:4000
• export HOST=127.0.0.1
• export PORT=3000
• export CASSANDRA_HOST=myhost
• sbt ui

Run Command Line UI node (at least one). Every instance need to have different PORT. It can be used to observe Cluster Singleton migration / failover.

• export SEED_NODE=akka.tcp://[email protected]:3000
• export SEED_NODE2=akka.tcp://[email protected]:4000
• export HOST=127.0.0.1
• export PORT=4000
• export CASSANDRA_HOST=myhost
• sbt cliui

Run Command Centers nodes (as many as you want, at least 1). Every instance need to have different PORT

• export SEED_NODE=akka.tcp://[email protected]:3000
• export SEED_NODE2=akka.tcp://[email protected]:4000
• export HOST=127.0.0.1
• export PORT=2501
• export CASSANDRA_HOST=myhost
• sbt cc

In UI choose click "Start" to begin invasion. When attack is running you can kill or restart Command centers node and watch what will happened.

Enjoy

How to run using a few computers

First build runnable application by sbt clean universal:packageBin. Result of build will be zip file missiledefend-x.y.z.zip in target/universal/ Use ./missiledefend-x.y.z/bin/missiledefend to run application (with arg ui for UI, cc for command center node. You have to also export following env variables:

• HOST - IP address used for communication
• PORT - Port number
• SEED_NODE - Akka seed node for example: akka.tcp://[email protected]:3000
• SEED_NODE2 - Second Akka seed node for example: akka.tcp://[email protected]:4000
• CASSANDRA_HOST - Contact point for Cassandra, default value is 127.0.0.1

Start nodes:

• missiledefend ui - UI - This node will host StatusKeeper actor (Cluster Singleton)
• missiledefend cliui - Command line UI - This node will host StatusKeeper actor (Cluster Singleton)
• missiledefend cc - Command Center to host defence towers
• missiledefend cc - Another Command Center on another machine

Scenarios to run

Power failure, JVM/System crash

• Start cluster
• Start game
• Kill one of "Command Center" JVM
• Wait for towers to be migrated and node to be disconnected (change color to red)
• Restart JVM
• Wait for node to be connected again

Network failure

• Start cluster with one Command Center node on remote machine
• Disconnect network on remote machine
• Wait for towers to be migrated and node to be disconnected (change color to red)
• Connect network again, node will not be connected again.
• Restart JVM, node will be connected again

Persistence crash

• Start cluster
• Start game
• Kill Cassandra
• Watch problems of Tower Actors (Sharding depends on persistence)
• Restart Cassandra
• Situation should go back to normal after while (if outage was short, less than ~30s)

Drop some traffic

• Start cluster with one Command Center node on remote machine
• Use "iptable" to drop 1%, 2%, 5%, 10% of traffic using iptables command:

  iptables -A INPUT -m statistic --mode random --probability 0.2 -j DROP 
  iptables -A OUTPUT -m statistic --mode random --probability 0.2 -j DROP 
  

• Watch how cluster behaves

Graceful leaving cluster

• Start cluster
• Start game
• Right click on one of command centers, select "Cluster leave"
• Watch migration time
• Restart node
• Run "Cluster leave again"

Detecting how many messages where lost

• Start cluster with 2 "Command Center" nodes

• Start game

• Click on one of the towers, it's name should be displayed.

• Kill "Command Center" on which selected tower is running

• Wait for towers to be migrated and node.

• Kill cluster

• On Ui under the tower will be displayed icon . Number under the icon shows how many messages are lost. In case the messages were delivered in wrong order, system also detect message lost.

  License

  Code is available under Apache Commons 2.0 License Icons are available under a Creative Commons Attribution 3.0 License. Most icons are downloaded from http://p.yusukekamiyamane.com/