So I was toying around with the load balancer trying to get familiar with it. I spun up a junk jenkins from the catalog to use as my destination and then created a load balancer in the same stack.
I edited Jenkins to not publish a port.
Finally, when I go to add the configuration I want to test into the load balancer:
And click save - it does not take and seems to ignore it.
However if I just enter ONLY the destination service it seems to publish the route to jenkins on port 80 (with no DNS logic of course)
I was able to add a load balancer with a hostname and request path in front of the jenkins service (from the catalog and upgraded it to remove the exposed 8080 port). I had no issues saving it.
Based on your image, if you added only the destination service (no host/request path), it will re-route the traffic to the exposed port 8080 of the container. The UI only shows the source port and not the target port. If you checked the haproxy.cfg on the load balancer, you’d be able to see that it’s targeting the 8080 port of the service.
To check the haproxy.cfg, you can exec into the shell of the load balancer and cat /etc/haproxy/haproxy.cfg. The backend should show the IP of the containers in the targetted service and port 8080.
I am using the latest version, I have continued testing and I have found that I can get DNS based routing to work if no /endpoint is used, but adding an endpoint seems to totally trash all rules.
If I have a chance I am going to recreate my problem setup and check inside to see if haproxy is getting the config or what config it is getting from Rancher.
