Scripts in this directory are necessary due to bugs (as of Q1-2017) in GKE that cause network load-balancer resources to become orphaned and unused but accruing charges against the account. These charges can be substantial. Before we realized what was happening (with the help of google support) we were seeing 40-50% of the monthly bill going to these network load-balancer resources.

• Usage
  • Configuration
  • Running the script
• Deployment
• Development & Testing

Configuration is handled through environment variables:

• PROJECT: The GCE project that we should operate on
• REGION: The region where GCE resources should be probed
• GKE_CLUSTER_NAME: The Kube cluster name for verifying network resources against
• KUBE_CONTEXT: the kube context to use when running kubectl commands - not needed when running inside a cluster.

Set the env variables and execute the script

PROJECT=myproject \
REGION=us-central1 \
GKE_CLUSTER_NAME=dev-01 \
KUBE_CONTEXT=gke_some-project_us-west1-a_dev01 \
./delete-orphaned-kube-network-load-balancers.sh

This script is derived from the similar script in the kubernetes github repo. That script deletes load-balancers that are pointing to nodes that no longer exist. This is helpful but does not cleanup all orphaned resources.

The delete-orphaned-kube-network-load-balancers.sh is written and maintained by Pantheon and offers a more complete cleanup function. It uses kubectl to get a list of public IP's assigned to active Services and then iterates through gcloud firewall-rules and forwarding-rules looking for IP's that are not in use by the cluster.

See the deployment-example.yaml file for an example Deployment.

It's just a simple shell script. All code should pass shellcheck linting (make test or make test-shell) and follow the Google Shell Style Guide.

make build-docker will build the docker container. make push will push it to quay.io. You can override the repo by setting the REGISTRY variable on the make task.