Accidently nuked half of some resources and broke DNS (yes, it is in fact always DNS). One of the first things I learned and is on a lot of guides for terraform is how `count` works. It’s one of the meta-arguments you can use with most resources, others are

``` depends_on count for_each provider / providers lifecycle ```

Here’s an example for count before I show my oops.

```hcl variable “names” { default = [“alice”, “bob”, “carol”] }

resource “aws_instance” “web” { count = length(var.names) ami = “ami-0c55b159cbfafe1f0” instance_type = “t3.micro”

tags = { Name = var.names[count.index] } } ```

Remove `“bob”` → `default = [“alice”, “carol”]`. You’ll see this in your terraform run of web[1] transitioning.

``` ~ Name = “bob” -> “carol” ```

carol shifts from index 2 → 1, so Terraform modifies the bob instance to become carol, and destroys the old carol. Two changes instead of one.

`count` is very quick and easy to use but honestly I avoided it. If I read of a feature that has to be used with extra considerations, I’d rather use the pattern that doesn’t let me screw up if my coffee has fully kicked in.

In my case, I added to middle of an array in a variable but the implementation logic was a count instead of a `for_each`.

Here’s an implementation that guarentess uniqness in the array to avoid collisions and doesn’t care about order.

```hcl

Before (vulnerable — index shuffle on any list change)#

resource “aws_instance” “web” { count = length(var.names) ami = “ami-0c55b159cbfafe1f0” instance_type = “t3.micro”

tags = { Name = var.names[count.index] } }

After (safe — each instance is an independent resource)#

resource “aws_instance” “web” { for_each = toset(var.instances) ami = “ami-0c55b159cbfafe1f0” instance_type = “t3.micro”

tags = { Name = each.key } } ```

Even better is that with the first method you access the resource like `aws_instance.web[0]` but with the 2nd you get a much more descriptive and assuring `aws_instance.web[“bob”]`