ingress_manual.md

First make your service type=NodePort

Then create an instance group in UI (console.cloud.google.com), with some pool of instances from one of your zones:

gcloud compute --project $PROJECT instance-groups create unmanaged $K8S_IG
gcloud compute --project $PROJECTinstance-groups unmanaged add-instances $K8S_IG --instances $NODE,$NODE_1...

Add the Service NodePort to the InstanceGroup:

gcloud compute --project $PROJECT instance-groups set-named-ports $K8S_G --named-ports svc1:$SVC1_NODE_PORT

Create a health check:

gcloud compute --project $PROJECT http-health-checks create "svc1-health-check" --port $SVC1_NODE_PORT

Create a backend service:

gcloud compute --project $PROJECT backend-services create "svc1-backends" --port $SVC1_NODE_PORT --protocol "HTTP" --http-health-check "svc1-health-check"

Add the backend to the IG:

gcloud compute --project $PROJECT backend-services add-backend "svc1-backends" --balancing-mode "UTILIZATION" --instance-group $K8S_IG

Create a firewall rule:

gcloud compute firewall-rules create allow-130-211-0-0-22  --source-ranges 130.211.0.0/22  --allow tcp:30000-32767

Where 30000-32767 is the nodeport range (you can replace it with just your nodeport) and that /22 is the range of ips for lb health checks (it's always that).

There are actually a bunch of other resources you need, but for the rest, using the ui is better. Now you should be able to navigate to the gce console and click on networking > http loadbalancing, and do what seems natural (url map, plug your backend in there etc).