mewld

mewld

A generic discord bot clusterer. Will be open source to help other large bot developers.

Mewld clusterer is rather generic and any bot should be able to adapt to it without any changes to mewld

What is a cluster

A cluster is a way to scale large discord bots by breaking them into seperate processes, each with a set of shards. mewld is an implementation of this system.

Terminology

• Cluster - A single process with a set of shards in it
• Cluster size - The amount of shards in a cluster
• Embedding - Embedding mewld into a Go project for increased customizability

How it works

• Shard count is retrieved using Get Gateway Bot. This is then used to create a ClusterMap assigning each cluster a set of shards based on per_cluster in config.yaml. These ClusterMap's are then used to make a InstanceList of Instance structs.

• First instance is then started, once this instance sends launch_next, the next instance is then launched until every cluster is launched, the last cluster should also send a launch_next once fully ready

• mewld also handles other tasks such as cluster management via its webui which also comes with (upcoming) support for application command permission configs (for private commands that should only be visible to specific roles in a support or staff server)

• mewld supports max_concurrency under the following limits. Firstly, the underlying bot must support launching several shards concurrently (as mewld does not actually start shards). Secondly, the number of shards that can actually be

Redis

Mewld uses redis for communication with clusters and for action logs. Action logs are stored as a Redis list under ${redis_channel_name}/actlogs.

Data Format:

JSON with the following structure (golang syntax to make updating docs simpler)

Scope     string         `json:"scope"`
Action    string         `json:"action"`
Args      map[string]any `json:"args,omitempty"`
CommandId string         `json:"command_id,omitempty"`
Output    any            `json:"output,omitempty"`
Data      map[string]any `json:"data,omitempty"` // Used in action logs

Operations

Diagnostics

Whenever a diagnostic is sent over the $CHANNEL channel, the cluster must respond with a diag (see Operations table) within 10 seconds. A diagnostic can be identified based on the existence of a diag key

FAQ

• Why are there so many PingCheckStop <- i.ClusterID?

Answer: To ensure that no erroneous ping checks are still running in background thus leading to random cluster death.