Unneeded health check created for kube-system:default-http-backend service

[tomoe]

When I create an Ingress that has default backends, the controller still creates a GCE healthcheck pointing to kube-system:default-http-backend service (default 404 server) even thought there's no chance of hitting the pod.

Usually, that doesn't do any harm, but possibly duplicated healthcheck probes may overload the l7-default-backend pod and causes the pod to restart with error messages because the resource setting for the container is very small (cpu: 10m, memory: 20M).

Now I think:

• it would be nice if HC gets created if it really is needed
• (not an issue of the controller, but) l7-default-backend resource setting may need to be increased, autoscalable, and/or configurable.

[rramkumar1]

@tomoe This issue should be fixed in GLBC v1.2.0. We will only create a backend service + health check for the system default backend if the user does not specify a default backend in their Ingress.

[rramkumar1]

/close

I still get that default BE. I even remove it manually and it forces the creation / link again.

[rramkumar1]

@AlbertoDominguezS What GKE cluster version are you using?

I would suggest filing a separate issue. It's much easier to track that way.

@rramkumar1 Cluster Version: 1.13.6-gke.5

[rramkumar1]

@AlbertoDominguezS Ok, would be useful to see your Ingress YAML so please file a separate issue. Thanks!