Skip to content

Latest commit

 

History

History
68 lines (42 loc) · 4.84 KB

migration-0.3.x.md

File metadata and controls

68 lines (42 loc) · 4.84 KB
page_title
Migration from 0.2.x to 0.3.x

Migration from 0.2.x to 0.3.x

Certain resources undergone changes in order to ensure consistency with REST API and standard expected Terraform behavior. You can upgrade provider with terraform init -upgrade.

provider

databricks_job

  • Rewrite spark_submit_parameters with spark_submit_task configuration block.
  • Rewrite python_file and python_parameters with spark_python_task configuration block.
  • Rewrite jar_uri, jar_main_class_name, and jar_parameters with spark_jar_task configuration block.
  • Rewrite notebook_path and notebook_base_parameters with notebook_task configuration block.
  • Rewrite library_jar, library_egg, library_whl, library_pypi, library_cran, and library_maven with library configuration block.

databricks_dbfs_file

  • Rename content to content_base64, as this closer represents actual data within the field and simplifies internal code reusability.
  • Remove overwrite attribute. Starting from v0.3.0 it behaves as if it is set to true.
  • Remove mkdirs attribute. Starting from v0.3.0 it behaves as if it is set to true.
  • Remove validate_remote_file attribute. Due to performance reasons, starting from v0.3.0 it doesn't fetch the contents of remote file to verify the checksum.
  • If you've relied on internal content_b64_md5 attribute, please remove it. Starting from v0.3.0 its behavior is internalized.

DBFS files would only be changed, if Terraform stage did change. This means that any manual changes to managed file won't be overwritten by Terraform, if there's no local change.

databricks_notebook

  • Rename content to content_base64, as this closer represents actual data within the field and simplifies internal code reusability.
  • Remove format attribute. Starting from v0.3.0 it behaves as if it is set to SOURCE.
  • Remove overwrite attribute. Starting from v0.3.0 it behaves as if it is set to true.
  • Remove mkdirs attributes. Starting from v0.3.0 it behaves as if it is set to true.

After changing the code, terraform apply would replace managed notebooks.

Notebook on Databricks workspace would only be changed, if Terraform stage did change. This means that any manual changes to managed notebook won't be overwritten by Terraform, if there's no local change to notebook sources. Notebooks are identified by their path, so changing notebook's name manually on the workspace and then applying Terraform state would result in creation of notebook from Terraform state.

databricks_cluster

  • Rewrite library_jar, library_egg, library_whl, library_pypi, library_cran, and library_maven with library configuration block.

databricks_instance_profile

  • Remove skip_validation from all databricks_instance_profile resources. In order to ensure consistency, all AWS EC2 profiles are now checked to work before returning state to main Terraform process.

databricks_mws_workspaces

  • Remove verify_workspace_runnning attribute from all databricks_mws_workspaces resources. All workspaces are verified to be running automatically as of this change.

databricks_instance_pool

  • Remove default_tags.

databricks_scim_user

  • This resource was removed as deprecated. Please rewrite using databricks_user.

databricks_scim_group

  • This resource was removed as deprecated. Please rewrite using databricks_group.

databricks_default_user_roles

  • This data source was removed as deprecated. Please use databricks_group data source for performing equivalent tasks.