Replace servers instead of patching them

How to do it

Build the image in CI with a tool such as Packer, or ship a container image. Start the new servers, wait for their health checks, then stop the old ones. A server changed by hand drifts away from the code, and nobody can rebuild it the same way. Replace each server at least every 30 days, so the newest updates are always baked in.

Example

resource "aws_launch_template" "web" {
  image_id      = data.aws_ami.web_latest.id
  instance_type = "t4g.small"
}

resource "aws_autoscaling_group" "web" {
  min_size = 2
  max_size = 6
  launch_template {
    id      = aws_launch_template.web.id
    version = aws_launch_template.web.latest_version
  }
  instance_refresh {
    strategy = "Rolling"
    preferences { min_healthy_percentage = 100 }
  }
}

How to check

No running server is older than 30 days. Every server runs the image ID of the latest build.