### Terraform Version ```shell Terraform v1.6.4 on linux_amd64 ``` ### Terraform Configuration Files N/A ### Debug Output N/A ### Expected Behavior **Note:** I originally thought this was a `cdktf` bug but it seems this issue is related to `terraform` directly. The expected behaviour and reproduction steps still used `cdktf` although it shouldn't be necessary to reproduce. --- I am running `cdktf destroy` via a python using `subprocess.run`. The exact command is ```python import subprocess subprocess.run(["cdktf", "destroy", "--auto-approve"], check=True) ``` When this is destroying instances, pressing `ctrl +c` should stop the destroy and leave the `terraform..tfstate` file with an accurate list of state. ### Actual Behavior The `terraform..tfstate` file is empty. ### Steps to Reproduce **Note:** The below instructions are one way to induce the issue but it is likely there are others. Using raw `terraform` it should be possible if the state is large and the `destroy` command is exited forcefully enough. --- 1. Start an instance with a lot of resources. For example many files. 2. Run `cdktf destroy` via python ```python import subprocess subprocess.run(["cdktf", "destroy", "--auto-approve"], check=True) ``` 4. When the resources are being destroyed, press `ctrl +c`. It seems like the `python` subprocess call is required for the state to be emptied permanently, however using a raw `cdktf destroy` I noticed that the state file is briefly emptied and then recreated. This seems dependent on the size of the state. I can reliably reproduce this with a ~5MB state file. ### Additional Context I believe this issue is related to the `TODO` comment here https://github.com/hashicorp/terraform/blob/1f9734619f953ecc7252d7b98a6d40d751b4ea1e/internal/states/statemgr/filesystem.go#L124C1-L129C18 With a large state file, an interrupt can happen between [clearing the file](https://github.com/hashicorp/terraform/blob/1f9734619f953ecc7252d7b98a6d40d751b4ea1e/internal/states/statemgr/filesystem.go#L194), and the updated state being completely [written](https://github.com/hashicorp/terraform/blob/1f9734619f953ecc7252d7b98a6d40d751b4ea1e/internal/states/statemgr/filesystem.go#L219). The solution is likely what the comment suggests. Write the state file to a temporary file and then do a "swap" to update the state file. This would ensure the state file is never in a "corrupted" state. ### References I originally reported this against under cdktf https://github.com/hashicorp/terraform-cdk/issues/3269