Asset Bundle deploy fails: terraform apply: failed to instantiate provider

dbph commented 1 month ago

Describe the issue

Every "bundle deploy" fails locally on my machine, but works perfectly, when I deploy with a azure-devops deployment pipeline.

Error message is "Error: failed to read schema for [my_job_name] in registry.terraform.io/databricks/databricks: failed to instantiate provider "registry.terraform.io/databricks/databricks" to obtain schema: Unrecognized remote plugin message: This usually means that the plugin is either invalid or simply needs to be recompiled to support the latest protocol."

After the process has failed, If I change the h1 checksum (only add one character) in .terraform.lock.hcl in the .databricks folder of the bundle and run "bundle deploy" for a second time, the deployment finishes successfully. The previous version of the h1 hash is then readded to the .terraform.lock.hcl file.

However, if I then run a "bundle deploy" again, It fails with the original error message.

Transforming the h1 hash again results in another successfull run... and so on.

Steps to reproduce the behavior

Create a new bundle with databricks bundle init default-python then run databricks bundle deploy then error.

Expected Behavior

bundle deploy results in "Deployment complete!"

Actual Behavior

bundle deploy results in "Error: failed to read schema for [my_job_name] in registry.terraform.io/databricks/databricks: failed to instantiate provider "registry.terraform.io/databricks/databricks" to obtain schema: Unrecognized remote plugin message: This usually means that the plugin is either invalid or simply needs to be recompiled to support the latest protocol."

OS and CLI version

Windows with Databricks CLI v0.218.1 but I've also tried several other versions.

andrewnester commented 1 month ago

Could you try to upgrade to the latest version (0.224.0), remove Terraform cache folder and see if the error persists?

dbph commented 1 month ago

Hi, thanks for your help!

I deleted folder C:\Users{my_user}\AppData\Roaming\terraform.d, deleted .databricks folder in the asset bundle folder and downloaded the latest Version of Databricks CLI:

databricks -v returns Databricks CLI v0.224.0
databricks bundle deploy unfortunately returns the same error as before.

Now the trick with changing the h1-hash of databricks provider in .terraform.lock.hcl. doesn't work anymore

pietern commented 1 month ago

@dbph Could you try running the Terraform provider directly as a binary?

If I do this on my machine, I get:

% .databricks/bundle/dev/terraform/.terraform/providers/registry.terraform.io/databricks/databricks/1.48.0/darwin_arm64/terraform-provider-databricks_v1.48.0
Databricks Terraform Provider

Version 1.48.0

https://registry.terraform.io/providers/databricks/databricks/latest/docs

This binary is a plugin. These are not meant to be executed directly.
Please execute the program that consumes these plugins, which will
load any plugins automatically

dbph commented 1 month ago

Hi, in the meantime, I reinstalled the previous version of the CLI (he one that is currently shipped in the vscode extension), therefore when executing the databricks terraform provider exe I get:

Databricks Terraform Provider Version 1.40.0 https://registry.terraform.io/providers/databricks/databricks/latest/docs This binary is a plugin. These are not meant to be executed directly. Please execute the program that consumes these plugins, which will load any plugins automatically

dbph commented 1 month ago

Hi,

anything elese, I could try?

Thanks :-)

andrewnester commented 1 month ago

It looks like something might have been cached somewhere and also your TF provider seems to be quite old hence the error. You sure you don't have any local installation of TF provider somewhere else on your machine?

dbph commented 1 month ago

I checked again. There is no local installation of the TF provider on my machine. The old TF provider was because I used the Databricks CLI shipped with the Databricks VS-Code Addin. I've tried again with Databricks CLI v0.224.1 resulting in TF Provider 1.49.1.

I still get the same error. If I delete terraform-provider-databricks_v1.49.1.exe in .databricks\bundle\dev\terraform.terraform\providers\registry.terraform.io\databricks\databricks\1.49.1\windows_amd64 in the bundle folder I want to deploy then the deployment finishes successfully. If I try it again - the same error. If I delete the file again then again it works once.

databricks / cli