Megane
A sharding manager for Discord bots. Megane distributes your shards across (logical) CPU cores. Megane uses Eris to interface with the Discord API. Based on Kurasuta.
Features
- Automatic clustering and sharding of your bot across CPU cores
- Central data store that all clusters can access and modify
- Simple and extensible way to get users, channels, and guilds from other clusters
- Run eval over IPC
- Restart individual clusters instead of the whole bot at once
- Create worker processes to interact with APIs or do other expensive or central tasks
- Automatic and extensible collection of statistics
Features to add before 1.0:
- (IPC) Optional messages in text channels
- Change cpuUsage recording to return diff since last check
Features to be added after 1.0:
- Rolling restart helper
- Re-sharding
- Improve IPC usability (need suggestions)
Considerations
Splitting your bot into multiple processes should not be done unless needed. Megane has some downsides compared to how a normal bot works. Make sure you consider these first:
- Developing will be more complicated due to the need to use ipc and lack of a complete local cache
- Fetching users, channels, guilds, or other structures from other processes will remove the ability to use methods without manually recreating them. Certain properties may also be missing after converting to plain objects
Using
Package name: @brussell98/megane
You can see working examples by browsing the examples
folder
Getting Started
You will have at least two main files:
- A file for your "master" process, which creates a new
ShardManager
. We will call this file "index.js" - A file for your "cluster" process (the worker), which extends
BaseClusterWorker
. We will call this file "bot.js"
In your index.js
file you should have some code similar to this:
const { isMaster } = require('cluster');
const { ShardManager } = require('@brussell98/megane');
const manager = new ShardManager({
path: __dirname + '/bot.js',
token: 'Your bot token'
});
manager.spawn(); // You should await this
if (isMaster) {
// Master process code here
manager.on('error', error => console.error(error)); // Not handling these errors will kill everything when any error is emitted
manager.on('debug', message => console.log(message));
manager.on('clusterReady', cluster => console.log(`Cluster ${cluster.id} ready`));
manager.once('allClustersReady', () => console.log('All clusters ready'));
}
This will create a new ShardManager
which will run bot.js
on separate processes. The worker file (bot.js
) must implement a BaseClusterWorker
. This will be demonstrated next.
Note the isMaster
block. Your index.js file will be run each time a worker is created. Any code you only want to run on the master process must check if it's running on the master process.
Next, your bot.js
file should implement a ClusterWorker, like so:
const { BaseClusterWorker } = require('@brussell98/megane');
module.exports = class BotWorker extends BaseClusterWorker {
constructor(manager) {
super(manager);
}
async launch() {
// Anything you want to run when the worker starts
// This is run after the IPC is initialized
await this.client.connect(); // Connect eris to Discord
}
}
Services
In many cases you will have tasks that are used by all clusters. One example is updating Twitch API data. You can create a service to handle this and it will run in its own process, accessible by all clusters. Add the following code to your index.js
file:
if (isMaster) {
// ...
manager.on('serviceReady', service => console.log(`Service ${service.name} ready`));
manager.registerService(__dirname + '/service.js', { name: 'example-service', timeout: 60e3 });
}
This will register a service named "example-service" and spawn a worker. The service worker is implemented similarly to a cluster worker.
Here is an example of what your service.js
file should look like:
const { BaseServiceWorker } = require('@brussell98/megane');
module.exports = class ServiceWorker extends BaseServiceWorker {
constructor(manager) {
super(manager);
}
async launch() {
// Anything you want to run when the worker starts
// This is run after the IPC is initialized
await this.sendReady(); // Required to be sent before `timeout`
}
// Handles SERVICE_COMMAND events
handleCommand(data, receptive) {
if (data.op === 'PING')
return receptive && this.asResponse('PONG');
if (receptive)
return this.asError(new Error('Unknown command'));
}
}
Your bot can then send commands like this:
const reply = await this.ipc.sendCommand('example-service', { op: 'PING' }, { receptive: true });
console.log(reply); // PONG
Diagram
Master:
index.js -> ShardManager -> Clusters
-> Services
-> MasterIPC
ClusterWorker:
index.js -> ShardManager -> bot.js -> ClusterWorkerIPC
ServiceWorker:
index.js -> ShardManager -> service.js -> ServiceWorkerIPC
Documentation
Refer to DOCUMENTATION.md
Changelog
Refer to CHANGELOG.md
Naming
This was created to be used for Mirai Bot for Discord. The bot is named after the anime character Mirai Kuriyama (栗山 未来), who notably wears red glasses [Megane (めがね)].