YakDriver
commented
3 years ago @jameswilsongrant Sorry this isn't working right. To be clear, when you say in step 1, to create a cluster in Terraform, step 2 will be performed in the same state as step 1? If they are in the same state, Terraform's dependency graph should make sure that delete for the asg, capacity provider, and template are called first.
There is an acceptance test that mirrors your scenario called TestAccAWSEcsCluster_SingleCapacityProvider. I've pieced together the configuration for the test. This test passes okay meaning that Terraform was able to create and destroy everything without timing out. If you see differences in the test configuration and yours that change how this behaves, let us know.
data "aws_ami" "test" {
most_recent = true
owners = ["amazon"]
filter {
name = "name"
values = ["amzn-ami-hvm-*-x86_64-gp2"]
}
}
data "aws_availability_zones" "available" {
state = "available"
filter {
name = "opt-in-status"
values = ["opt-in-not-required"]
}
}
resource "aws_launch_template" "test" {
image_id = data.aws_ami.test.id
instance_type = "t3.micro"
name = "yakluster"
}
resource "aws_autoscaling_group" "test" {
availability_zones = data.aws_availability_zones.available.names
desired_capacity = 0
max_size = 0
min_size = 0
name = "yakluster"
launch_template {
id = aws_launch_template.test.id
}
tags = [
{
key = "foo"
value = "bar"
propagate_at_launch = true
},
]
}
resource "aws_ecs_capacity_provider" "test" {
name = "yakluster"
auto_scaling_group_provider {
auto_scaling_group_arn = aws_autoscaling_group.test.arn
}
}
resource "aws_ecs_cluster" "test" {
name = "yakluster"
capacity_providers = [aws_ecs_capacity_provider.test.name]
default_capacity_provider_strategy {
base = 1
capacity_provider = aws_ecs_capacity_provider.test.name
weight = 1
}
}
odormond
richardgavel
jaffel-lc
mavin-audacia
github-actions[bot]
There's an ordering issue with resource destruction when ecs has ec2 capacity:
You'll end up like below. All the tasks/services are deleted but the capacity provider and ASG will still be there. There will still be whatever ec2s the ASG is creating attached to the cluster. This will then sit until timeout. Deleting the ASG at any point will move it along.
I'm guessing there has to be inspection from ecs -> capacity provider -> asg to destroy this in the right order (destroying the asg will terminate the instances as well). Seems like a bit of an edge case though.