⛔Broadcast Rate Limiting
Rate limiting helps you throttle client events, backend events, or HTTP REST API calls for read endpoints (like /channels
) at the app level.
Choosing a rate limiter driver depends on the architecture the server is deployed in. For local, single-instance servers, the default local driver is sufficient. For multi-node deployments, a Redis server is typically required. If you choose to use Redis to store rate limiting data, please consult our documentation on connecting to Redis.
Environment Variables
Name | Default | Possible values | Description |
---|---|---|---|
|
|
| The driver used for rate limiting. |
|
|
| Only for the |
local
- Rate limiting data is stored in-memory and is lost upon server exit.redis
- Rate limiting data is centralized in Redis. This driver is recommended when deploying multi-node soketi configurations.cluster
- Rate limiting data is centralized within the master instance, that is elected. This driver is recommended when deploying multi-threaded soketi configurations (i.e. PM2)
Last updated