Return 503 by default when no endpoint is available

[maxlaverse]

Return 503 by default when no endpoint is available

What this PR does / why we need it:
This PRs changes the backend of a location to be empty per default instead of "upstream-default-backend". During template rendering, if the backend of a location is empty, a 503 is returned to the client.

Which issue this PR fixes
Currently, if no endpoint is available for an upstream, the backends of locations using that upstream will be set to the default-backend.

This means that a client might get a 404under heavy load when no end-points are available. This sound wrong has clients interpret 404 as a valid response meaning the resource doesn't exist. I think that instead of a 404, the answer should be a 503 Service Unavailable (The server is currently unable to handle the request).

The definition of the default backend from the documentation says it's a service [for] handling all url paths and hosts the nginx controller doesn't understand.

I think that perfectly fine and would like not to use it anymore for upstreams without endpoints, which the Nginx Controller knows about but are simply temporarily unavailable.

Special notes for your reviewer
I didn't tested the behavior when using a custom-default-backend, but I expect it to be unchanged by this PR.

[k8s-reviewable]

This change is 

[coveralls]

Coverage remained the same at 33.492% when pulling 1062340 on maxlaverse:return_503_by_default into 8b94f46 on kubernetes:master.

[aledbf]

/lgtm

[aledbf]

@maxlaverse thanks!

[joshbranham]

I am not sure if this is relevant here, but I am trying to utilize the custom-default-backend to serve custom 50x error pages, as well as a custom 404. The 404 page works as expected, however when the nginx controller has no valid endpoints for a rule, it seems to just display its own 503 error page. Does this PR reflect that I need to custom the 503 page inside each ingress-controller, versus allowing the default backend to handle it as stated here: https://github.com/kubernetes/ingress-nginx/blob/master/docs/examples/customization/custom-errors/README.md

[joshbranham]

I solved my above issue by adding the custom config map value of custom-http-errors: 404,502,503,504. This seemed a bit unclear given the docs on custom-errors