Backends healthchecks and expected operation

[caquino]

Hi,

I've been trying to troubleshoot some issues I'm having while deploying ingress-gce, but my lack of experience with a working installation limits sometimes my expectations.

After deploying it to my cluster, I can see a new ingress created:

Name:             ai-cloud-production-ingress
Namespace:        default
Address:          <REDACTED>
Default backend:  default-http-backend:80 (<REDACTED>:8080)
Rules:
  Host  Path  Backends
  ----  ----  --------
  *     
           api:80 (<REDACTED>:80,<REDACTED>:80,<REDACTED>:80)
Annotations:
  ingress.kubernetes.io/backends:               {"k8s-be-30167--01177f64ec764488":"HEALTHY","k8s-be-32416--01177f64ec764488":"UNHEALTHY"}
  ingress.kubernetes.io/forwarding-rule:        k8s-fw-default-ai-cloud-production-ingress--01177f64ec764488
  ingress.kubernetes.io/https-target-proxy:     k8s-tps-default-ai-cloud-production-ingress--01177f64ec764488
  ingress.kubernetes.io/url-map:                k8s-um-default-ai-cloud-production-ingress--01177f64ec764488
  kubernetes.io/ingress.global-static-ip-name:  ai-cloud-global-production
  ingress.gcp.kubernetes.io/pre-shared-cert:    ai-cloud-production-ssl
  ingress.kubernetes.io/https-forwarding-rule:  k8s-fws-default-ai-cloud-production-ingress--01177f64ec764488
  ingress.kubernetes.io/ssl-cert:               ai-cloud-production-ssl
  ingress.kubernetes.io/target-proxy:           k8s-tp-default-ai-cloud-production-ingress--01177f64ec764488
Events:                                         <none>

As can be seen on the describe ingress, one of my backends is listed as UNHEALTHY.

After verifying the backend configuration via cloud console, I was able to identify that the health check assigned to the backend has the incorrect URI.

• Is it expected to have 2 backends for a service in a single cluster?

• How can I troubleshoot the URI used for health check as it's incorrect?

[caquino]

After reading the FAQ looks like some of my questions where answered, please correct me if I'm wrong:

• There are 2 backends because one is the default backend.

Regarding the health check, the FAQ points to an example that don't exist so it's hard to compare to what I have on my Deployment definition.

I have a Liveness probe configured for my service, which I assume ingress-gce would use as source for the health check:

Liveness:     http-get http://:80/healthz delay=5s timeout=5s period=10s #success=1 #failure=2

[caquino]

Just found the answer for my question on #241