WebSocket Connection Failure in Self-Hosted Supabase on EC2

[brookbdt]

Issue: WebSocket Connection Failure in Self-Hosted Supabase on EC2

Description

When attempting to connect to Supabase Realtime through a self-hosted instance on EC2, the WebSocket connection fails.

Environment

• Deployment: Self-hosted on AWS EC2
• Domain: {supabasePublicUrl}
• Connection URL: wss://{supabasePublicUrl}/realtime/v1/websocket
• Client Parameters:
  • apikey: [ANONKEY]
  • eventsPerSecond: 2
  • vsn: 1.0.0

Steps to Reproduce

1. Set up self-hosted Supabase on EC2
2. Attempt to establish WebSocket connection to the Realtime service
3. Connection fails immediately

Expected Behavior

• WebSocket connection should establish successfully
• Client should maintain stable connection to Realtime service

Additional Context

• The server logs show multiple GET requests to the WebSocket endpoint
• GET /realtime/v1/websocket returning 404
• Client receives a connection failure without specific error details

Questions

1. Are there specific configuration requirements for self-hosted Realtime service on EC2?
2. Are there known networking/firewall considerations for WebSocket connections in this setup?
3. How can we enable more detailed error logging to diagnose the connection failure?

[jsmueller7]

This is probably caused by a cookie that is too large. It doesn't occur if you're using a local account. But with SSO I've seen this. Even if you set MAX_HEADER_LENGTH to 8192, the cookie can cause a problem with the ws:// connection.

A temporary fix in this case is to let Kong remove the cookie with a request-transformer.

[filipecabaco]

You can also configure it with MAX_HEADER_LENGTH

realtime/config/runtime.exs

Line 71 in afb2a1b

max_header_value_length: String.to_integer(System.get_env("MAX_HEADER_LENGTH") || "4096")

[colabcolibri]

Any news about it?