Additional EBS volumes are re-attached when an instance is recreated

[BernhardBerbuir]

When tainting an EC2 Instance which belongs to a group of instances with externally attached volumes then all externally attached volumes are re-attached.

How to reproduce:

• create multiple EC2 instances with this module
• create for each instance an additional EBS volume and an attachment (see comment )
• taint one of the instances
• execute terraform apply

As a result Terraform re-creates the volume attachments for all instances

aws_volume_attachment.data[0] must be replaced
-/+ resource "aws_volume_attachment" "data" {
...
  instance_id = "i-09223024b168b857c" -> (known after apply) # forces replacement
...
aws_volume_attachment.data[1] must be replaced
-/+ resource "aws_volume_attachment" "data" {
...
  instance_id = "i-07144ca24f14bc773" -> (known after apply) # forces replacement
...

I assume that this happens because the additional EBS volumes are created and attached outside of the module.
I could create a pull request which creates aws_ebs_volume and aws_volume_attachment inside the module, if it's ok.

[BernhardBerbuir]

I have done some further analysis of the bug.
The statefile contains the ebs_block_device two times:

1. there are separate aws_ebs_volume and aws_volume_attachment (as expected)
2. each aws_instance contains also an ebs_block_device (instead of an empty array)

=> the aws_instance "hijacks" the separately created and attached ebs_block_device.

[antonbabenko]

@BernhardBerbuir I think you can try to use for_each instead of count for this situation and loop through ids instead of count indexes. This will solve half of the issues you are describing.

Check this for more details - https://www.terraform.io/docs/configuration/resources.html#when-to-use-for_each-instead-of-count

Unfortunately, it is not possible to specify variables inside ignore_changes, so we can't have ebs_block_device ignored for some but not for all.

There is nothing we can do in the module, so I am closing this issue.

[github-actions]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues. If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.