NoProxy in some self host environments

I've recently deployed SourceGraph with the Code Host GitHub. The environment where it was delpoyed uses injected proxies for the internal applications to reach the internet; this includes http, https and no_proxy. I had quite some issues for a while, having to track the traffic to see what went wrong.

Specifically, there was a problem in the repo-updater pod that caused other pods to fail as well; the error message found in the logs being: [...] error="invalid character '<' looking for beginning of value". Furthermore, the gitserver-0 pod failed once the repo-updater had been fixed.

Solution:
The problem in my case was that the proxy was routing internal traffic using the cluster internal DNS to the internet, which of course returned errors. In this specific case, an error page that returned HTML content, which caused the go code to fail.
When setting proxies for the pods, the gitserver-0 and repo-updater pods require a no_proxy/NO_PROXY envvar with the following values: localhost,sourcegraph-frontend-internal,gitserver-0.gitserver. This is assuming that your http/https proxies forward all other traffic elsewhere.

Given that this has caused quite a headache while fixing,this might be worth adding somewhere in your documentation, hopefully helping out other PoCs.

not sure if it's a similar issue, but at least the error message and the use of a proxy was identical in https://github.com/sourcegraph/sourcegraph/issues/2360

L’issue identifie les pods gitserver-0 et repo-updater ainsi que les valeurs NO_PROXY requises localhost, sourcegraph-frontend-internal et gitserver-0.gitserver. Trouvez l’entrée pertinente de la documentation Helm concernant la configuration du proxy des pods, documentez ces paramètres et le contexte de l’échec, et vérifiez que les instructions couvrent les deux pods.