Why it matters
An idle load balancer does nothing but cost money, and it is easy to miss among dozens of active ones.How nOps finds it
nOps reads the last 90 days of billing for each load balancer. One is flagged when its hourly charge was billed on at least 85 of those days and in the last 2 days, and no line on the load balancer recorded any usage (processed bytes, capacity units or data transfer) in the whole 90 days. The saving is the hourly charge billed in the last 30 days, and must be at least $1.Before you start
Permission to delete the load balancer (elasticloadbalancing:DeleteLoadBalancer). Find out who owns it, and whether it was created by Kubernetes, CloudFormation or Terraform.
How to implement
- Check the load balancer’s tags. If it has
elbv2.k8s.aws/cluster,kubernetes.io/cluster/...oraws:cloudformation:stack-nametags, remove it from its owner (the Kubernetes Ingress or Service, or the stack) instead of deleting it directly. - Confirm it is not a standby or disaster-recovery load balancer, and that no DNS record points to it.
- Turn off deletion protection if it is on.
- Delete it: EC2 → Load Balancers → Actions → Delete, or
aws elbv2 delete-load-balancer --load-balancer-arn ARN(aws elb delete-load-balancer --load-balancer-name NAMEfor a Classic Load Balancer). - Remove it from Terraform or CloudFormation so it is not recreated.
Risks and considerations
- Deleting a load balancer cannot be undone. A new one gets a new DNS name.
- A load balancer deleted by hand that Kubernetes or a stack manages is recreated, or leaves that tool out of sync.
- Target groups are not deleted with the load balancer; remove them separately if they are no longer used.
FAQs
Could a load balancer with no traffic still be needed?
Could a load balancer with no traffic still be needed?
Yes, for example a disaster-recovery standby. That is why nOps asks you to confirm before deleting.