Add gcp example to expose sourcegraph publically

[michaellzc]

I was testing helm experience on GCP and happened to feel like giving Ingress a try. It turned out to be a lot harder than I was expecting. Also, GCP docs is better than other vendors, but still, it's hard to troubleshoot when you're not on their expected user path.

Our frontend deployment uses a different port 6060 (other than the one severs user traffic 3080) to perform health check. GCP ingress controller (or load balancer) requires the present of health check (readinessProbe) in order to work, but things get complicated when the healthcheck port is different than the port published from the svc. Therefore, it requires a special BackendConfig CR to tell ingress controller where to check. Otherwise, it will just fall back to check / at 80, and the load balancer ends up reporting backed is unhealthy cuz there's nothing listening at 80.

Side note, I don't know why we went down the road of deploying our own nginx ingress controller on GCP with NLB, but using platform built-int just feel like the better choice to me.

Is this something we want to demonstrate to users? This is related to the point we brought up in the planning doc about creating reference architecture and how to support them. For this example, we can potentially go even further including steps of provisioning TLS cert + DNS to get sourcegraph working properly on GKE.

Related: kubernetes/ingress-gce#42

Checklist

• Follow the manual testing process

Test plan

Deploy at https://console.cloud.google.com/kubernetes/list/overview?project=michael-test-03&supportedpurview=project

http://34.120.16.199/ this works ;)

[michaellzc]

Current dependencies on/for this PR:

• main
  • PR Add gcp example to expose sourcegraph publically #69 (this PR)

This comment was auto-generated by Graphite.

[daxmc99]

LGTM on this approach. Looks like we are essentially adding more docs here as this in the example folder

[michaellzc]

LGTM on this approach. Looks like we are essentially adding more docs here as this in the example folder

docsite should be the single source of truth, but we also need a place to store all these sample override files and some local context is also nice. I am still working out the balance about what to keep in the helm repo and what go to the docsite, here's something WIP https://github.com/sourcegraph/sourcegraph/pull/32463.

The plan for helm page on docsite is to keep it as a one-pager, so we can break it up or move sections around when the new docsite IA is ready (an initiativeive led by Jeff, Director of Technical Writing). Maybe we will move all example to docsite or figure out a way to reference them from docsite after GA.