v0.12: Resource timeouts being ignored on destroy operations

[alexng-canuck]

Terraform Version

terraform version is 0.12.0
provider is based on plugin SDK version 0.12.0-rc1

Terraform Configuration Files

resource oci_core_instance instance {
   timeouts {
    create = "60m"
    delete = "60m"
    update = "60m"
  }
}

Debug Output

Crash Output

Expected Behavior

On Delete calls, the provider calls into ResourceData.Timeout(schema.TimeoutDelete) and expects to get 60 minutes time value back.

This works in v0.11.7 plugin SDKs

Actual Behavior

The ResourceData.Timeout(schema.TimeoutDelete) is returning a default timeout value instead (20m).

Steps to Reproduce

This was found on running acceptance tests, but a simple terraform destroy should be able to repro this.

Additional Context

On Create calls, the timeout values are correct. This seems to only affect Delete operations.

Timeout values that are specified in the resource schema are also ignored even if timeouts block is removed from config. For example, if my resource has the following schema it will still return the default of 20m.

&schema.Resource{
		Timeouts: &schema.ResourceTimeout{
			Create: &OneHour,
			Update: &OneHour,
			Delete: &OneHour,
		},

References

[danawillow]

@jbardin, is it possible there's another place where a fix needs to be made? I vendored 0.12.2 into the Google provider yesterday, but I'm still seeing tests fail due to delete timeouts- here's an example: https://ci-oss.hashicorp.engineering/repository/download/GoogleCloud_ProviderGoogleCloudGoogleProject/61683:id/debug-google-770e78c-TestAccContainerCluster_errorCleanDanglingCluster.log (the delete timeout for container clusters is 30m, but the test timed out after 20m)

[jbardin]

Thanks @danawillow! I've replicated the case where individual resource provided defaults are being lost at some point in their lifecycle.

[jbardin]

Thanks again @danawillow! I merged #21814 which catches the case I think you were hitting.