5
$ terraform -version

Terraform v0.12.21
+ provider.aws v2.34.0

Given a terraform configuration for one set of single aws_instance, aws_ebs_volume and aws_volume_attachment resources, each named, say,foo;

How would one terraform destroy -target=aws_instance.foo without terraform also destroying the aws_ebs_volume.foo as the resulting plan insists that it will?

  • The aws console reports Delete on termination for the volume's block device to be false;
  • I've tried to terraform destroy -target=aws_volume_attachement.foo first;
    ... no-change.
  • I've tried removing the aws_volume_attachment from the configuration, terraform apply it and then terraform destroy -target=aws_instance.foo;
    ... didn't help either

Somehow terraform really wants to destroy the aws_ebs_volume resource together with the aws_instance resource.

I, otoh, would very much like to preserve that aws_ebs_volume for later attachment to a re-created aws_instance...

Can it be done?

cueedee
  • 151
  • 6

2 Answers2

2

This may be what you are doing already (in which case this sounds like a bug), yet it seems just based on the limited information in the question that you may be forgetting a step: Run plan -destroy -target=... -target=... first. The terraform destroy command will try to operate on changes listed in the .plan file.

The following example assumes a Terraform module named foo, with 3 EBS volumes and attachments named as: dev-sdb, dev-sdf, dev-sdg. All resources are assumed to have count = local.enabled ? 1 : 0. So this is the more advanced case you may encounter where each resource would have to be targeted with the array syntax [0].

  1. First create a targeted plan file with terraform plan -destroy -target=.... Do this before trying to run terraform destroy

    terraform plan -destroy \
               -target=module.foo.aws_volume_attachment.dev-sdb-attachments[0] \
               -target=module.foo.aws_volume_attachment.dev-sdf-attachments[0] \
               -target=module.foo.aws_volume_attachment.dev-sdg-attachments[0] \
               -target=module.foo.aws_instance.foo[0]  \
               -target=module.foo.aws_iam_instance_profile.foo[0] \
               -out=intermediate.plan

  2. Run terraform destroy, optionally also using -target=... for safety precautions if you wish:

    terraform destroy -target=module.foo.aws_volume_attachment.dev-sdb-attachments[0] \
                  -target=module.foo.aws_volume_attachment.dev-sdf-attachments[0] \
                  -target=module.foo.aws_volume_attachment.dev-sdg-attachments[0] \
                  -target=module.foo.aws_instance.foo[0]  \
                  -target=module.foo.aws_iam_instance_profile.foo[0] \
                  intermediate.plan

If you need this to be unattended for automation, add -auto-approve flag to the terraform destroy command. Otherwise, it assumes an interactive terminal session and will prompt before destroying any resources.

TrinitronX
  • 178
  • 5
0

A workaround to your issue is to create a snapshot of the volume(s) before deleting them. The snapshot will live on after the volume is deleted and can be recreated if necessary.

After you no longer need an Amazon EBS volume, you can delete it. After deletion, its data is gone and the volume can't be attached to any instance. However, before deletion, you can store a snapshot of the volume, which you can use to re-create the volume later.

Reference

EBS Deleting Volume

kenlukas
  • 672
  • 7
  • 15
  • That is more or less what I ended up doing before reaching out to stackexchange, but this is an involved workaround at best. – cueedee Mar 01 '20 at 17:48