Jonathon Williams, submitted 25 October 2011

• Python, on Linux (tested on Ubuntu Linux 11.04)
• Twisted Python (available in Ubuntu's apt package manager as python-twisted; this should already be installed in the Assignment 1 ISISCloud image).

1. Run an instance, which will be the machine you'll run the launch script on. Give this machine a public IP address.

2. Copy the scripts in this directory to the remote instance. From the current directory (the one with this README file in it) on your machine, run:

   scp -i <path to your key file> *.py *.sh ubuntu@<instance ip address>:~/
   

   replacing <path to your key file> and <instance ip address> with the appropriate values.

3. Copy your key file and novarc to the home directory of the remote instance. Make sure it's named in the form <key-name>.pem, where <key-name> is the name of your key as used with the -k option to euca-run-instances (i.e. jwilliams2.pem if the name of your key were jwilliams2).

4. Connect to the remote instance via SSH:

   ssh -i <path to your key file> ubuntu@<instance ip address>
   

5. Source novarc:

   source novarc
   

6. Set the name of your key file:

   export KEY_NAME=<key-name>
   

   where <key-name> is the name of your key, without the .pem suffix (i.e. export KEY_NAME=jwilliams).

7. Run the script, and follow any on-screen instructions:

   ./launch_test.sh
   

• ISISCloud is still a bit unreliable... if an instance fails to run or SSH into the instance doesn't work, the script will fail. You should be able to press Ctrl+C to stop the script, then kill the instances it created and restart the script to try again.
• Log files are left in the same directory as the launch script: these contain the standard output of the test scripts to verify that they're working correctly.

Each program needs to be started in a particular order:

1. Fault Manager
2. Server(s)
3. Client

Wait a few seconds between starting each component to give things a chance to start running and stabilize.

From a command line:

python FaultManager.py <port>

replacing <port> with a port number of your choice (needs to not be in use, and should be greater than 1024 to run as a regular user). Remember this number; it is needed by the client and servers.

From a command line:

python Server.py <local-ip> <local-port> <fault-manager-ip> <fault-manager-port>

Required parameters:

• local-ip: The IP address of the machine this server is running on. This should be the IP where the fault manager can reach this replica.
• local-port: An arbitrary port number of your choice (greater than 1024). If more than one server or the fault manager is running on the same machine, these values must all be different.
• fault-manager-ip: The IP address of the fault manager.
• fault-manager-port: The port number of the fault manager.

A program to consume an arbitrary amount of CPU time is also provided. This can be run from the command line:

python UseCpu.py <percentage>

where percentage is a rough amount of CPU to consume.

The client runs in a loop, sending a request to the active server replica approximately every ten seconds, then printing the results (or an error message if no replicas are available).

From a command line:

python Client.py <fault-manager-ip> <fault-manager-port>

Required parameters:

• fault-manager-ip: The IP address of the fault manager.
• fault-manager-port: The port number of the fault manager.