Classic Load Balancers should have connection draining enabled

How to fix it

In the console

EC2 → Load balancers → the load balancer → Attributes → Traffic configuration

Enable connection draining

Selected

Timeout (draining interval)

Your longest normal request, 1 to 3,600 seconds (default 300); longer also slows instance replacement

With the AWS CLI

aws elb modify-load-balancer-attributes --load-balancer-name <load-balancer-name> --load-balancer-attributes ConnectionDraining={Enabled=true,Timeout=300} --region <region>

With Terraform

resource "aws_elb" "this" {
  connection_draining         = true
  connection_draining_timeout = 300

  listener {
    lb_port           = 80
    lb_protocol       = "http"
    instance_port     = 80
    instance_protocol = "http"
  }
}