Looking at the aws console there are 5 load balancers for community cluster. Master and app nodes both have 2, one of which is classic and one is application load balancer, then there is one more load balancer (internal).

This is how DNS was configured:

• community.skunkhenry.com -> master nodes, application load balancer
• *.community.skunkhenry.com -> app nodes, classic load balancer

I've tried to reproduce the issue with websockets in qe aws account. I used application loadbalancer there and everything worked well. So only thing I've changed now was change the DNS record for *.community.skunkhenry.com to point to the prepared application load balancer for application nodes.

I've tested this setup and websockets work for me. Also tested data sync with subscriptions.

But I'm not sure if pointing *.community.skunkhenry.com to classic lb was not intentional. Comparing with application lb I was not able to find anything special about this load balancer. Everything seems to work fine now so far.

Michael Nairn, David Kirwan, Jesse Sarnovsky is this change I made ok? Or is it possible this change will cause different issues in future? If you think we should use the classic lb, just let me know, I can revert the change.