Getting started

When to shard

Before you dive into this section, please note that sharding may not be necessary for you. Sharding is only required at 2,500 guilds—at that point, Discord will not allow your bot to login without sharding. With that in mind, you should consider this when your bot is around 2,000 guilds, which should be enough time to get this working. Contrary to popular belief, sharding itself is very simple. It can be complicated depending on your bot's needs, however. If your bot is in a total of 2,000 or more servers, then please continue with this guide. Otherwise, it may be a good idea to wait until then.

How does sharding work?

As an application grows large, a developer may find it necessary to split their process to run parallel to maximize efficiency. On a much larger scale of things, the developer might notice their process slow down, amongst other problems. Check out the official Discord documentation on the topic.open in new window

WARNING

This guide only explains the basics of sharding using the built-in ShardingManager, which can run shards as separate processes or threads on a single machine. If you need to scale beyond that (e.g., running shards on multiple machines/containers), you can still do it with discord.js by passing appropriate options to the Client constructor. Nevertheless, you will be on your own regarding managing shards and sharing information between them.

TIP

Apart from ShardingManager, discord.js also supports a sharding mode known as Internal sharding. Internal sharding creates multiple websocket connections from the same process, and does not require major code changes. To enable it, simply pass shards: 'auto' as ClientOptions to the Client constructor. However, internal sharding is not ideal for bigger bots due to high memory usage of the single main process and will not be further discussed in this guide.

Sharding file

First, you'll need to have a file that you'll be launching from now on, rather than your original index.js file. It's highly recommended renaming that to bot.js and naming this new file to index.js instead. Copy & paste the following snippet into your new index.js file.

const { ShardingManager } = require('discord.js');

const manager = new ShardingManager('./bot.js', { token: 'your-token-goes-here' });

manager.on('shardCreate', shard => console.log(`Launched shard ${shard.id}`));

manager.spawn();
1
2
3
4
5
6
7

The above code utilizes the discord.js sharding manager to spawn the recommended amount of shards for your bot. The recommended amount should be approximately 1,000 guilds per shard. Note that you have to attach the event listener to shardCreate before calling .spawn() to prevent a race condition possibly preventing shard 0 from logging the successful launch. Even though you provide the token here, you will still need to send it over to the main bot file in client.login(), so don't forget to do that.

TIP

You can find the methods available for the ShardingManager class hereopen in new window. Though, you may not be making much use of this section, unlike the next feature we will explore, which you may learn about by clicking this link.

Getting started

You will most likely have to change some code to get your newly sharded bot to work. If your bot is very basic, then you're in luck! We assume you probably have some form of a stats command, by which you can quickly view your bot's statistics, such as its server count. We will use it as an example that needs to adapt to running with shards.

In this code, you likely have the snippet client.guilds.cache.size, which counts the number of cached guilds attached to that client. Since sharding will launch multiple processes, each process (each shard) will now have its subset collection of guilds it is responsible for. This means that your original code will not function as you might expect.

Here is some sample code for a stats command, without sharding taken into consideration:

// bot.js
const { Client, Intents } = require('discord.js');

const client = new Client({ intents: [Intents.FLAGS.GUILDS, Intents.FLAGS.GUILD_MESSAGES] });

client.on('interactionCreate', interaction => {
	if (!interaction.isCommand()) return;

	const { commandName: command } = interaction;

	if (command === 'stats') {
		return interation.reply(`Server count: ${client.guilds.cache.size}.`);
	}
});

client.login('your-token-goes-here');
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

Let's say your bot is in a total of 3,600 guilds. Using the recommended shard count, you might end up at four shards, each containing approximately 900 guilds. If a guild is on a specific shard (shard #2, for example) and receives this command, the guild count will be close to 900, which is not the "correct" number of guilds for your bot. Let's take a look at how to fix that.

FetchClientValues

One of the most common sharding utility methods you'll be using is Shard#fetchClientValuesopen in new window. This method retrieves a property on the Client object of all shards.

Take the following snippet of code:

client.shard.fetchClientValues('guilds.cache.size').then(console.log);
1

If you run it, you will notice an output like [898, 901, 900, 901]. You will be correct in assuming that that's the total number of guilds per shard stored in an array in the Promise. This probably isn't the ideal output for guild count, so let's use Array.reduce()open in new window to provide a better output.

TIP

It's highly recommended for you to visit the documentationopen in new window to understand how the reduce() method works, as you will probably find great use of it in sharding.

In this case, this method iterates through the array and adds each current value to the total amount:

client.shard.fetchClientValues('guilds.cache.size')
	.then(results => {
		console.log(`${results.reduce((acc, guildCount) => acc + guildCount, 0)} total guilds`);
	})
	.catch(console.error);
1
2
3
4
5

While it's a bit unattractive to have more nesting in your commands, it is necessary when not using async/await. Now, the code at the top should look something like the below:

client.on('interactionCreate', interaction => {
	// ...
	if (command === 'stats') {
		return client.shard.fetchClientValues('guilds.cache.size')
			.then(results => {
				return interaction.reply(`Server count: ${results.reduce((acc, guildCount) => acc + guildCount, 0)}`);
			})
			.catch(console.error);
	}
});



 
 
 
 
 


1
2
3
4
5
6
7
8
9
10

BroadcastEval

Next, check out another handy sharding method known as Shard#broadcastEvalopen in new window. This method makes all of the shards evaluate a given method, which recieves a client and a context argument. The client argument refers to the Client object of the shard evaluating it. You can read about the context argument here.

client.shard
	.broadcastEval(c => c.guilds.cache.reduce((acc, guild) => acc + guild.memberCount, 0))
	.then(console.log);
1
2
3

This will run the code given to broadcastEval on each shard and return the results to the Promise as an array, once again. You should see something like [9001, 16658, 13337, 15687] logged. The code sent to each shard adds up the memberCount property of every guild that shard is handling and returns it, so each shard's total guild member count. Of course, if you want to total up the member count of every shard, you can do the same thing again on the Promise results.

client.shard
	.broadcastEval(c => c.guilds.cache.reduce((acc, guild) => acc + guild.memberCount, 0))
	.then(results => {
		return interaction.reply(`Total member count: ${results.reduce((acc, memberCount) => acc + memberCount, 0)}`);
	})
	.catch(console.error);
1
2
3
4
5
6

Putting them together

You'd likely want to output both pieces of information in the stats command. You can combine these two results with Promise.all()open in new window:

const promises = [
	client.shard.fetchClientValues('guilds.cache.size'),
	client.shard.broadcastEval(c => c.guilds.cache.reduce((acc, guild) => acc + guild.memberCount, 0)),
];

Promise.all(promises)
	.then(results => {
		const totalGuilds = results[0].reduce((acc, guildCount) => acc + guildCount, 0);
		const totalMembers = results[1].reduce((acc, memberCount) => acc + memberCount, 0);
		return interaction.reply(`Server count: ${totalGuilds}\nMember count: ${totalMembers}`);
	})
	.catch(console.error);
1
2
3
4
5
6
7
8
9
10
11
12

Promise.all() runs every Promise you pass inside an array in parallel and waits for each to finish before returning their results simultaneously. The result is an array that corresponds with the array of Promises you pass–so the first result element will be from the first Promise. With that, your stats command should look something like this:

client.on('interactionCreate', interaction => {
	// ...
	if (command === 'stats') {
		const promises = [
			client.shard.fetchClientValues('guilds.cache.size'),
			client.shard.broadcastEval(c => c.guilds.cache.reduce((acc, guild) => acc + guild.memberCount, 0)),
		];

		return Promise.all(promises)
			.then(results => {
				const totalGuilds = results[0].reduce((acc, guildCount) => acc + guildCount, 0);
				const totalMembers = results[1].reduce((acc, memberCount) => acc + memberCount, 0);
				return interaction.reply(`Server count: ${totalGuilds}\nMember count: ${totalMembers}`);
			})
			.catch(console.error);
	}
});



 
 
 
 
 
 
 
 
 
 
 
 


1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17

The next section contains additional changes you might want to consider, which you may learn about by clicking this link.

Resulting code

If you want to compare your code to the code we've constructed so far, you can review it over on the GitHub repository here open in new window.