Terraform state mv destroys resources

Question

I'm performing refactoring in a few Terraform files. One of the tasks is renaming resources from kebab-case to snake_case.

To prevent the destruction and recreation of said resources, I used terraform state mv. Now, for some reason I'm hoping to understand here, it still destroys resources. I see 2 issues:

1- The IDs get computed again.

2- The references to variables are taken as literals.

Example:

-/+ aws_volume_attachment.att_ebs_caldat_axon_apps (new resource required)
      id:                                                 "vai-4287143552" => <computed> (forces new resource)
      device_name:                                        "/dev/xvdb" => "/dev/xvdb"
      force_detach:                                       "true" => "true"
      instance_id:                                        "i-0ca294d44635d3ace" => "${module.instance_axon.instance_id}" (forces new resource)
      volume_id:                                          "vol-0298f5247bb2aa312" => "${aws_ebs_volume.ebs_caldat_axon_apps.id}" (forces new resource)

I'm using Terraform 0.11.14

The command for moving the state for this resource was terraform state mv aws_volume_attachment.att-ebs-caldat-axon-apps aws_volume_attachment.att_ebs_caldat_axon_apps

I don't know what I'm missing. Any help is appreciated.

Can you follow this doc , its well written ryaneschinger.com/blog/terraform-state-move , you need to also think of dependencies and work bottom to top approach .Also to comment anything we need to know your whole config file — Surya Prakash Patel
So according to that article I should've done the state moves in a specific order taking resource dependencies into account... I honestly thought if I did them all at once it would work... Now I guess I need to figure out how to rollback the state since I'm in remote and moreover it is a production env. — Sanne Tuin
I suggest you to backup your state file ( as it is prod ) and try editing states — Surya Prakash Patel
There is one state file backup per (apparently) each state mv that I performed. Would it be safe to do a state push of the oldest backup to revert all the performed moves? It requires a -force flag since "the destination state has a higher serial number" — Sanne Tuin

Sanne Tuin Sanne Tuin · Accepted Answer · 2020-05-15T06:49:44

Turns out terraform will recreate the EBS resources if you change either their volume or attachment names even if you perform a state mv. I don't fully understand the reason for this but the same behavior is not encountered when doing the same with other resources i.e. security groups.

Thank you.

Terraform state mv destroys resources

1 Answers