I've upgraded a couple of Google Container Engine clusters from Kubernetes version 1.2.4 to 1.3.5, and according to the UI the upgrade process is finished. However, hitting the corresponding Web servers (which are behind L7 load balancers AKA GLBCs) results in error 502 even though it's been about half an hour since the upgrade process finished.
How do I fix/troubleshoot this? I figure it's GLBC that's having trouble, as it's typically returning 502 while transitioning, but I have no idea how to troubleshoot it.
Seems as if the ingresses of my different clusters were clashing with each other after the Kubernetes upgrade due to ingress UIDs no longer working or some such, at least that's the current theory. Reinstalling L7 ingresses for all clusters with UIDs reconfigured fixed it.