add deletion endpoint

Author: CrysR

src/webapp/databricks.py

@@ -66,7 +66,6 @@ def setup_new_inst(self, inst_name: str) -> None:
         )
         db_inst_name = databricksify_inst_name(inst_name)
         cat_name = databricks_vars["CATALOG_NAME"]
-        print(cat_name)
         for medallion in medallion_levels:
             w.schemas.create(name=f"{db_inst_name}_{medallion}", catalog_name=cat_name)
         # Create a managed volume in the bronze schema for internal pipeline data.
@@ -143,3 +142,40 @@ def run_pdp_inference(
             },
         )
         return DatabricksInferenceRunResponse(job_run_id=run_job.response.run_id)
+
+    def delete_inst(self, inst_name: str) -> None:
+        db_inst_name = databricksify_inst_name(inst_name)
+        cat_name = databricks_vars["CATALOG_NAME"]
+        w = WorkspaceClient(
+            host=databricks_vars["DATABRICKS_HOST_URL"],
+            # This should still be cloud run, since it's cloud run triggering the databricks
+            # this account needs to exist on Databricks as well and needs to have permissions.
+            google_service_account=gcs_vars["GCP_SERVICE_ACCOUNT_EMAIL"],
+        )
+        # Delete the managed volume.
+        w.volumes.delete(name=f"{cat_name}.{db_inst_name}_bronze.bronze_volume")
+        w.volumes.delete(name=f"{cat_name}.{db_inst_name}_silver.silver_volume")
+        w.volumes.delete(name=f"{cat_name}.{db_inst_name}_gold.gold_volume")
+
+        # Delete the MLflow model.
+        # TODO how to handle deleting all models?
+        """
+        model_name = "latest_enrollment_model"
+        new_institution_model_uri = f"{cat_name}.{db_inst_name}_gold.{model_name}"
+        mlflow_client.delete_registered_model(name=new_institution_model_uri)
+        """
+
+        # Delete tables and schemas for each medallion level.
+        for medallion in medallion_levels:
+            all_tables = [
+                table.name
+                for table in w.tables.list(
+                    catalog_name=cat_name,
+                    schema_name=f"{db_inst_name}_{medallion}",
+                )
+            ]
+            for table in all_tables:
+                w.tables.delete(
+                    full_name=f"{cat_name}.{db_inst_name}_{medallion}.{table}"
+                )
+            w.schemas.delete(full_name=f"{cat_name}.{db_inst_name}_{medallion}")

src/webapp/routers/institutions.py

@@ -231,6 +231,52 @@ def create_institution(
     }
 
 
+# TODO: add tests
+@router.delete("/institutions/{inst_id}", response_model=None)
+def delete_inst(
+    inst_id: str,
+    current_user: Annotated[BaseUser, Depends(get_current_active_user)],
+    sql_session: Annotated[Session, Depends(get_session)],
+    storage_control: Annotated[StorageControl, Depends(StorageControl)],
+    databricks_control: Annotated[DatabricksControl, Depends(DatabricksControl)],
+) -> Any:
+    """Delete an existing institution.
+
+    Only available to Datakinders.
+
+    Args:
+        current_user: the user making the request.
+    """
+    if not current_user.is_datakinder():
+        raise HTTPException(
+            status_code=status.HTTP_401_UNAUTHORIZED,
+            detail="Not authorized to delete an institution.",
+        )
+
+    local_session.set(sql_session)
+    local_session.get().execute(
+        delete(InstTable).where(InstTable.id == str_to_uuid(inst_id))
+    )
+    local_session.get().commit()
+    # Delete GCS bucket
+    bucket_name = get_external_bucket_name_from_uuid(query_result[0][0].id)
+    try:
+        storage_control.delete_bucket(bucket_name)
+    except ValueError as e:
+        raise HTTPException(
+            status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
+            detail="Storage bucket deletion failed:" + str(e),
+        )
+    # Delete all databricks managed pieces.
+    try:
+        databricks_control.delete_inst(query_result[0][0].name)
+    except Exception as e:
+        raise HTTPException(
+            status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
+            detail="Databricks deletion failed:" + str(e),
+        )
+
+
 # All other API transactions require the UUID as an identifier, this allows the UUID lookup by human readable name.
 @router.get("/institutions/name/{inst_name}", response_model=Institution)
 def read_inst_name(