### Terraform Version ```shell 1.0.11, 1.8.5, 1.10.5 (doesn't occur in Terraform 0.12.31) ``` ### Terraform Configuration Files I have a custom Terraform resource and Terraform provider, let's call it `custom_resource_dbinstance`. In schema, it has multiple properties, like `database_id`, `engine_version`, `db_version`. I treat `database_id` and `engine_version` as inputs, and `db_version` property as output - it has `Computed: true` set in schema configuration, and I update its value in ReadContext, CreateContext and UpdateContext. Then, with the use of the custom provider, I have the Terraform configuration as below. In `main.tf`, I have: ```terraform resource "custom_resource_dbinstance" "db" { database_id = "agatest" engine_version = "14.0.0" } ``` In `outputs.tf`, I have: ```terraform output "DBVersion" { value = custom_resource_dbinstance.db.db_version } output "DatabaseEngineVersion" { value = custom_resource_dbinstance.db.engine_version } ``` After apply, `DBVersion` output is set to `14.12` (so far, as expected, computed correctly and has the same value as `db_version` property in tf state), and `DatabaseEngineVersion` is set to `14.0.0`. Then, I change the Terraform configuration below and perform terraform apply. In `main.tf`, I have: ```terraform resource "custom_resource_dbinstance" "db" { database_id = "agatest" engine_version = "14.1.0" } ``` In `outputs.tf`, I have: ```terraform output "DBVersion" { value = custom_resource_dbinstance.db.db_version } output "DatabaseEngineVersion" { value = custom_resource_dbinstance.db.engine_version } ``` Underneath, my computation changes `db_version` property in tf state to `14.15`. However, after apply, `DBVersion` output is still set to `14.12` (I expected it to be set to `14.15`, the same as `db_version` property in tf state), and `DatabaseEngineVersion` is set to `14.1.0` (ok). ### Debug Output ``` ...debug output, or link to a gist... ``` ### Expected Behavior After the second apply, I would expect `DBVersion` output value to be `14.15`, the same as `db_version` property value in tf state. This is how Terraform 0.12.31 behaved for me (I recently upgraded to 1.8.5). ### Actual Behavior After the second apply, the `DBVersion` output value is `14.12`, which is the same as the pervious `db_version` property value in tf state. Only when I execute third `terraform apply`, Terraform sees the difference and tries to update the output value. ``` Changes to Outputs: ~ DBVersion = "14.12" -> "14.15" ``` It seems that the output values are determined at the start of `terraform apply`, not at the end, when my `custom_resource_dbinstance` resource is updated. ### Steps to Reproduce 1. `terraform init` 2. `terraform apply` 3. Change `engine_version` input value. 4. `terraform apply` 5. `custom_resource_dbinstance.db.db_version` computed value changes and is visible in terraform state, but the output value based on `custom_resource_dbinstance.db.db_version` is still the same as after point 2. ### Additional Context _No response_ ### References _No response_ ### Generative AI / LLM assisted development? ChatGPT 3.5. My question: "Terraform 1.0 doesn't seem to treat computed properties of a resource as dependencies for the output. What would you suggest to do so that the output value is updated correctly after "terraform apply"?" Meaningful parts of the response: 1. Direct Reference: Ensure that your output is directly referencing the computed property. If the computed property itself changes, this should trigger an update to the output. ```terraform resource "aws_instance" "example" { # resource configuration } output "computed_property" { value = aws_instance.example.some_computed_property } ``` 2. Custom Null Resource: Create a null_resource that triggers when you expect changes, and reference this in your output. This can force Terraform to reevaluate the output when the null_resource changes. ```terraform resource "null_resource" "trigger" { triggers = { always_run = "${timestamp()}" } } output "computed_property" { value = aws_instance.example.some_computed_property depends_on = [null_resource.trigger] } ``` 3. Use of depends_on : If your output relies on a resource or module that might not directly change but could affect the output, you can use depends_on to force the output to wait for certain resources to complete. ```terraform output "example_output" { value = module.some_module.some_output depends_on = [aws_instance.example] } ``` None of the above worked.