“Provider produced inconsistent final plan”

[devopsix]

When applying a Terraform configuration which includes a herokux_app_container_release resource Terraform reports this error:

What is the nature of the issue?

• Bug
• Feature
• Enhancement

Terraform Version / Provider Versions

Affected Resource

• herokux_app_container_release

Terraform Configuration Files

This is an excerpt of the configuration. The actual configuration is a lot larger and unfortunately some parts cannot be disclosed without significant redaction.

resource "herokux_app_container_release" "app" {
  app_id = heroku_app.app.id
  image_id = data.herokux_registry_image.app.digest
  process_type = var.process_type
}

Expected Behavior

terraform apply succeeds and creates/updates the resource.

Actual Behavior

See error message above.

Steps to Reproduce

Please list the steps required to reproduce the issue, for example:

1. terraform init
2. terraform plan -lock=true -input=false -var-file=../images.tfvars.json -out=my.tfplan
3. terraform apply -lock=true -input=false -auto-approve my.tfplan

Important Factoids

We have recently updated from herokux v0.33.0 to v1.2.0. But we have had multiple successful runs of terraform apply since then. Also, other instances of this resource contained in the same configuration can be deployed successfully.

State contents of the “bad” resource:

The same configuration contains more instances of the herokux_app_container_release resource. A noticable difference is that their state does not have the timeouts attribute:

HashiCorp's help center article on “Provider Produced Inconsistent Results” suggests deleting the resource and re-running terraform apply. We have removed the resource from the remote state and afterwards were able to successfully run terraform apply for this configuration.

[davidji99]

Thanks for the bug report. Will take a look.