Add a project tag in sliderule-base.pkr.hcl to capture costs for the prov-sys

[cugarteblair]

inside the default_tags. tags= section
Project = "cluster-${var.cluster_name}"

`locals {
terraform-git-repo = "sliderule-cluster"
}

provider "aws" {
region = var.aws_region
default_tags {
tags = {
Owner = "SlideRule"
Project = "cluster-${var.cluster_name}"
terraform-base-path = replace(path.cwd,
"/^.*?(${local.terraform-git-repo}\/)/", "$1")
cost-grouping = "${var.cluster_name}"
}
}
}`

[jpswinski]

@cugarteblair - I went to add the project tag to the packer file, and I realized that there is no concept of a cluster for the packer build. We use the same AMI across all cluster and all types of instances. I think we want to keep it that way because it greatly simplifies managing it. Also, for security reasons, we never want to go back to an old AMI, or use anything other than the latest.

[cugarteblair]

JP,It seems to me that the AMI and project tag are independent issues. Correct?Carlos CarlosOn Aug 22, 2023, at 7:50 AM, JP Swinski ***@***.***> wrote: @cugarteblair - I went to add the project tag to the packer file, and I realized that there is no concept of a cluster for the packer build. We use the same AMI across all cluster and all types of instances. I think we want to keep it that way because it greatly simplifies managing it. Also, for security reasons, we never want to go back to an old AMI, or use anything other than the latest. —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you were mentioned.Message ID: ***@***.***>

[cugarteblair]

Is the issue that the packer is built independent of instantiating the EC2?

[jpswinski]

Yes, that's right. We maintain only three AMI's: "v3", "latest", and "unstable". In practice, v3 and latest are the only ones I keep up-to-date. We don't keep fully specified versions or old versions, because as soon as an AMI needs to be updated, it has to be updated across the board for every cluster.

[cugarteblair]

from chatgpt...
In AWS, there's a distinction between EC2 instance tags and Auto Scaling Group (ASG) tags. When you're using an Auto Scaling Group in Terraform, if you want the tags to propagate to the instances launched by that ASG, you must define them within the aws_autoscaling_group resource and ensure that propagate_at_launch is set to true for each tag.

The default_tags section under the provider "aws" block in your Terraform configuration will ensure that all AWS resources created by Terraform will get the defined tags. However, it won't ensure that the tags propagate to the instances launched by an ASG.

To ensure that the tags defined in the provider "aws" block get propagated to the EC2 instances launched by the ASG, you need to add those tags explicitly to the aws_autoscaling_group resource.

Here's how you can modify the aws_autoscaling_group block from your sliderule-asg.tf file:

resource "aws_autoscaling_group" "sliderule-cluster" {
  # ... (other properties)

  tag {
    key                 = "Name"
    value               = "${var.cluster_name}-node"
    propagate_at_launch = true
  }

  tag {
    key                 = "Owner"
    value               = "SlideRule"
    propagate_at_launch = true
  }

  tag {
    key                 = "Project"
    value               = "cluster-${var.cluster_name}"
    propagate_at_launch = true
  }

  tag {
    key                 = "terraform-base-path"
    value               = replace(path.cwd, "/^.*?(${local.terraform-git-repo}\\/)/", "$1")
    propagate_at_launch = true
  }

  tag {
    key                 = "cost-grouping"
    value               = "${var.cluster_name}"
    propagate_at_launch = true
  }
}

This will ensure that the tags you've defined in the provider "aws" block get propagated to the EC2 instances launched by the ASG.