ActionCable With Elastic Beanstalk and Classic Load Balancer 499 NGINX Code

Question

I am trying to use ActionCable and Turbo Stream in my Rails 7.0.4 app where the WebSocket connection to /cable is being closed unexpectedly in the production environment.

Ruby 3.0.5

64bit Amazon Linux 2/3.6.3

Implementing this Chat with Turbo Streams and OpenAI

https://gist.github.com/alexrudall/cb5ee1e109353ef358adb4e66631799d

This issue does not occur in development everything works perfectly there.

Redis Cable Config

production:
  adapter: redis
  url: <%= "redis://#{ENV.fetch('REDIS_HOST', 'localhost')}:6379/1"%>
  channel_prefix: fist_bump_production

REDIS_HOST is the AWS Elasticache service endpoint and have tested the endpoint works for Sidekiq.

First I had a 404 issue in production with /cable but then I added this to the config.

location /cable {
  proxy_pass http://my_app;
  proxy_http_version 1.1;
  proxy_set_header Upgrade websocket;
  proxy_set_header Connection "upgrade";
  proxy_set_header X-Real-IP $remote_addr;
  proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
}

No more 404 but then I started getting a 499 and need to refresh the page to see the updates.

I also see this in the console.

Have tried this in the config to permit all urls.

config.action_cable.allowed_request_origins = [ %r{http://*}, %r{https://*} ]

• I have increased stickiness on the Classic Load Balancer.
• Changed listeners to TCP on both 80 and 443.
• Read about 3 pages worth of Google results.
• tried to get help from AWS "Experts" (and I had to explain what Rails was)

No real help. Any suggestions on how to get this thing working in Beanstalk?

Ane

Answer 1

The 499 error code seems to point to a client losing connection with the nginx reverse proxy. I looked for you in the AWS elastic beanstalk documentation for reverse proxies. In the example config I see that AWS recommends adding timeouts.

Try adding this to your nginx conf:

      keepalive_timeout     60;

This would keep the connection open for 60 seconds as described in the nginx documentation;

Maybe this aswell:

      client_header_timeout 60;
      client_body_timeout   60;

Hope this helps.

Chronon

Thanks Max I ended up giving up after multiple days of trying this and actually moved to a different way of deploying directly on AWS. Of course everything worked perfectly once I did that. I am not sure if that setting is the same thing as the stickiness setting in the the ELB which did not work at all either. – Ane