Infrastructure/Terraform packages (presumably) not deleting since Feb 1

Up until today (2/1/2022) I have a terraform module repo in particular that is referenced by another terraform repo. The module is built automatically via GitLab CI on using shared runners and placed in the infrastructure registry. This has worked well for months. Since we haven’t released yet, I generally delete the module before triggering a new job so that I don’t have to increment the module version numbers in there referencing files (GitLab CI doesn’t overwrite terraform packages so you have to delete first).

Starting this morning around 9am Eastern time, I deleted the package (and nothing is listed in the infrastructure registry, but when my deploy process executes via curl, I get an error stating “Package version already exists.” Now, if I increment the module version, it deploys, but now I have to update every other place it’s referenced and it’s a pain.

The upload is executed with:

curl --silent --header "JOB-TOKEN: $CI_JOB_TOKEN" --upload-file ${file} "${CI_API_V4_URL}/projects/${CI_PROJECT_ID}/packages/terraform/modules/${MODULE_NAME}/${system}/${VERSION}/file"

edit: I should have mentioned that it appears that when I delete that it is deleting from their database but gitlab probably has a separate storage bucket for the terraform modules and the module isn’t being deleted from there.

Has anyone experienced this today? Any insights into this? I was hoping it would be resolved but it hasn’t yet. I’m worried it is somehow connected to yesterday’s outage, but I don’t have any proof of that.


Yes I am seeing the same problem. I delete a package using the terraform registry UI and it disappears. I also confirm its deletion using the API /packages?private-token=&package_type=terraform. All seems fine, however when I try to upload a replacement, I get the ‘Package version already exists’ message (HTTP status code of 403). One difference for my situation is that I only started doing this since the OP noticed the problem, so I can’t say whether this is new behaviour or not. Any help welcome :slight_smile:

I reached out to support and this has been fixed.