If your bot's shards are multi-processed, then a good choice is to use the [Gateway Queue]. The Gateway Queue is a lightweight but powerful application that you can host to queue shards across all of your processes.
The Gateway Queue is library and language agnostic, it's an HTTP server that you call whenever a shard needs to reconnect to the gateway.
When one of your shards disconnects and needs to perform a full reconnect, it needs to start a new session with the gateway. If you have multiple processes managing shards, you may not have communication between these processes. The problems start to happen when multiple shards from these processes try to reconnect at the same time, all but 1 will get rate-limited.
This is because there's a 5 second ratelimit between new sessions. By sending an HTTP request to the Gateway Queue, it will ratelimit the requests and "stall" them, responding with a friendly message once that shard can reconnect:
{
"message": "You're free to connect now! :)"
}The API is pretty minimal. It's just an HTTP request:
reqwest::get("http://gateway-queue").await?;No headers, body, or particular method need to be set, they're all ignored. The request will get a response once the request has gone through the queue.
If the bot has access to bucketed identify, you will have to make slight
configuration changes. Firstly you will need to set the DISCORD_TOKEN
environment variable so that it is able to fetch the remaining identifies and what
the max_concurrency is. Secondly, you will need to set the query parameter
shard to the ID of the identifying shard.
reqwest::get("http://gateway-queue?shard=42").await?;If you're using Docker, you can use the prebuilt Docker images from Github's container registry.
$ docker run -itd -e HOST=0.0.0.0 -e PORT=5000 ghcr.io/twilight-rs/gateway-queueIf you're not, you can compile it via Cargo:
$ cargo build --release
$ HOST=0.0.0.0 PORT=5000 ./target/release/twilight-gateway-queueHOST and PORT are the only two environment variables.