Real-time sockets for isitchristmas.com

This app is the socket server for flag/mouse streaming on isitchristmas.com. This server does not run isitchristmas.com -- for that code, see github.com/isitchristmas/web.

This service is meant to be run on a number of parallel nodes, e.g. 30 Heroku dynos or Nodejitsu drones or whatever. It's okay to run on a weak server, and each service can handle something like ~50 active users.

Setup

• Install Redis.

• Install Node.

• Copy the template config file:

cp config.js.example config.js

• Run the app at at http://localhost:3000:

node app.js

Deploying to Heroku

The recommended development/deployment path for Heroku is:

• Copy the template environment to .env and fill it out:

cp env.example .env

• Test out by running the app with foreman, with the port you want:

foreman start -p 3000

• (Recommended) Install the Heroku config plugin and sync your .env to the server:

heroku config:push -o

There's a Procfile in this repo already that will run node app.js.

Control Server

Every node connects to a Redis server, and the administrator with access to the Redis server can issue commands that get sent to all server nodes.

Server nodes can then send commands and config changes down to all connected users.

Commands

An admin can publish to a command channel to issue commands:

publish command refresh

Use :: to separate arguments:

publish command "blast:Testing the emergency broadcast system."

Commands include:

• refresh - Refresh browsers, using window.location. Useful for making client-side code deploys take effect.
• reconnect - Cause clients to disconnect and reconnect to the WebSockets endpoint. Useful for making server-side code deploys take effect.
• blast - Cause a message to appear in users' developer consoles.

Config changes

An admin can publish to either the server or client channel to affect "live" configuration.

server changes will be broadcast to every server and only saved there. client changes will be broadcast to every server and saved, and then broadcast down from each server to all connected clients.

Whether chat is enabled is only controlled server-side. Enable chat:

publish server chat:true

Changes to the mouse rate limit need to also be seen client-side. Rate limit mouse events to one every 50ms (20 fps):

publish client mouse_rate:50

The current state of "live" configuration is given to each client upon connecting to the server, so everyone should stay more or less in sync.

Server config values include:

• chat - Enable the chat system with true, turn it off with anything else.

Client config values include:

• mouse_rate - Rate limit, in milliseconds, for mouse events. 50ms = 20 "frames per second", and is a good value for high traffic use.
• heartbeat_interval - Interval, in milliseconds, where connected clients should "heartbeat" in to tell the server they're still around.
• death_interval - Time, in milliseconds, to wait for a heartbeat until a connected user is forcibly disconnected. Suggested value: 60000 (1 minute).
• ghost_duration - Time, in milliseconds, for a "ghost" flag to stick around until it disappears. (Defaults to 2000.)
• ghost_max - Number of "ghost" flags someone can have on screen at a given time. (Defaults to 10.)

License

Released under the MIT License.

--