Terraform project to demonstrate a gke/networking issue.

A private GKE cluster created using the following ip ranges,
creates a non operational gke cluster due to some core services be unreachable.

Issues noticed

• In the google cloud web console, under kubernetes engine show system workloads.

• When a new gke cluster is using this set of ipranges.
  Specifically when the container range is 172.17.0.0/16 the kube-dns services goes into a crash loop, then causing other services to crash after.

# main.tf, module "vpc"
ip_cidr_range      = "172.31.0.0/23"
service_ip_range   = "172.31.4.0/23"
container_ip_range = "172.17.0.0/16"

• Container logs for those crashing workloads indicate a network time out to/from the service ip range.

• The same module works fine when using a different set of iprages.

# main.tf, module "vpc"
ip_cidr_range      = "172.31.0.0/23"
service_ip_range   = "172.31.4.0/23"
container_ip_range = "172.16.0.0/16"

How to use this repository

• Clone the repo and modify the local values in the main.tf

locals {
  project = "< enter the gcp project name >"
  region  = "australia-southeast1"
  name    = "subnet-anomaly"
}

• Apply the configuration from the root folder.

terraform init
terraform plan
terraform apply

• Destroy

terraform destroy