We have a 6 node mixed OS UCP Swarm. 3 Linux Nodes and 3 Windows nodes.
We are using the UCP Layer 7 load balancing with the default nginx interlock extension over the default ports of 8080 and 8443.
The ucp-interlock-proxy only runs on the three linux nodes which we have setup as the swarm managers.
We also have an external HAproxy load balancer that does SSL termination and port translation.
Whenever a request comes into the HAproxy LB over 443, the traffic is load balanced among the 3 linux nodes running the ucp-interlock-proxy service over port 8080.
The ucp-interlock-proxy service is set with a publish mode of ingress.
The problem we’re running into is that we can deploy linux services and use the interlock proxy and everything works perfectly. When we deploy it we’ll give it the following labels, the port label is always the same but we’ll change the hosts per service.
This works perfectly with the linux services.
But when we try to do deploy windows services and do the same thing, same port different hostname, we get a 504 Gateway Time-out on haproxy, and if I examine the logs of the interlock proxy container running nginx these requests to windows nodes will have a status of 499, which is the haproxy hitting its timeout limit.
Does anyone have any idea why this would be happening?