distributed-control-panel

______  _____ ______
|  _  \/  __ \| ___ \
| | | || /  \/| |_/ /
| | | || |    |  __/
| |/ / | \__/\| |
|___/   \____/\_|

distributed-control-panel (henceforth DCP) is a python-based control panel to launch, control, and view distributed applications across machines. This is done via providing an abstraction over instances of the distributed application, and over available servers, to provide a central Python shell for all interactions with the application's instances.

Usecase

Building, testing, and showing demos of distributed p2p applications is a painful job. DCP attemps to change this. The idea is to provide an abstraction over all the instances of the your distributed application, and all the servers it is being run on.

DCP does this by launching an HTTP process on each of your servers (it uses SSH to authenticate on each of the servers the first time). This process now knows how to launch instances of your binary on each of the server, as well as how to stop them.

Furthermore, you get easy access to logs of each instance, full control over their CLI arguments, as well as other utility features.

All this is done inside an interactive Python shell, which means you can interact with the outputs in whatever way you like, within the comfort of Python.

A (potentially outdated) list of useful variables available at runtime:

• nodes
• servers
• logs
• shutdown_all_servers()
• setup_all_servers()
• run_binaries_on_all_nodes()
• add_new_node()
• stop_binaries_on_all_nodes()
• Tmux.ls(?server)

Running

./dcp.sh --help
./dcp.sh --cfg=config.json --nodes=3 --edges=5

Notes

• This requires the unbuffer program to be installed. On ubuntu, it can be installed by sudo apt-get install expect.
• The list of available variables is displayed on startup as well.
• All processes are run inside tmux sessions.
• The scripts are written for python 3.
• Each node is an instance of the Node class, and each server is an instance of the Server class.
• The classes Node and Server provide many more utility functions, which you can look into. This should allow boundless scripting freedom. Documentation for the same will be available in the future.
• The configuration allows you to provide a python function which outputs the cmd arguments of each node, given an instance of that node (type Node). See config.json for an example.