denik · Nov 5, 2026 · Nov 5, 2026 · Nov 5, 2026 · Nov 6, 2026
@@ -0,0 +1 @@
+* Rename a registered model in place when its `name` changes, keeping its versions, instead of deleting and recreating it. Registered models inside a renamed catalog or schema now follow it to the new name; recreating them failed because they already exist there. ([#7379](https://github.com/databricks/cli/pull/7379))
@@ -2,18 +2,17 @@
 === Replay the prior terraform deployment (captured from a real terraform run)
 >>> replay_tfstate.py terraform-requests.json terraform.tfstate
 
-=== Rename the model (an immutable id field) and deploy on direct
+=== Rename the model (an updatable id field) and deploy on direct
 >>> DATABRICKS_BUNDLE_ENGINE=direct [CLI] bundle deploy
-Warn: migration to direct: registered_models.my_model: name differs between config ("themodel-[UNIQUE_NAME]-renamed") and terraform state ("themodel-[UNIQUE_NAME]"); this resource will be recreated.
+Warn: migration to direct: registered_models.my_model: name differs between config ("themodel-[UNIQUE_NAME]-renamed") and terraform state ("themodel-[UNIQUE_NAME]"); this resource will be renamed.
 Notice: migrating your bundle to direct deployment engine (https://github.com/databricks/cli/issues/6765).
 Uploading bundle files to /Workspace/Users/[USERNAME]/.bundle/[UNIQUE_NAME]/files...
 Migrated 2 resources to direct deployment engine.
-Recreated registered_models.my_model
+Updated registered_models.my_model
 Files: 2 uploaded, 0 deleted
-Resources: 1 created, 0 changed, 1 deleted, 1 unchanged
+Resources: 0 created, 1 changed, 0 deleted, 1 unchanged
 
 >>> print_migration_telemetry
-direct_migrate_recreate_planned true
 direct_migrated_via_env true
 
 === Redeploy on direct: the model now matches config, so the plan is a clean no-op

@@ -10,12 +10,10 @@ title "Replay the prior terraform deployment (captured from a real terraform run
 trace replay_tfstate.py terraform-requests.json terraform.tfstate
 rm -f "$OUT_REQUESTS"
 
-title "Rename the model (an immutable id field) and deploy on direct"
-# name is a provided id field on registered models, so a rename recreates. reconcileIDFields
-# seeds the deployed name into the converted state, so the migration's plan sees a real
-# recreate. Unlike schemas/volumes, registered models are not in the destructive-approval
-# group (a model recreate loses no data), so the deploy is approved without --auto-approve:
-# the migration commits and the recreate applies on direct.
+title "Rename the model (an updatable id field) and deploy on direct"
+# name is an updatable id field: the model is renamed in place (UpdateWithID), keeping its
+# versions. reconcileIDFields seeds the deployed name into the converted state, so the
+# migration's plan sees a rename rather than snapshotting the new name as already applied.
 update_file.py databricks.yml "themodel-$UNIQUE_NAME" "themodel-$UNIQUE_NAME-renamed"
 trace DATABRICKS_BUNDLE_ENGINE=direct $CLI bundle deploy
 trace print_migration_telemetry

@@ -2,7 +2,7 @@ export DATABRICKS_BUNDLE_ENGINE=
 
 # Testserver-only: changing volume_type to EXTERNAL needs a storage_location backed by a
 # registered external location on a real workspace, so this is not Cloud-enabled. The
-# recreate class it covers is verified on cloud by schema-catalog-change / registered-model-rename.
+# recreate class it covers is verified on cloud by schema-catalog-change / registered-model-catalog-change.
 
 # volume_type is both a provided id field and a recreate_on_changes field. reconcileIDFields
 # records the deployed value, so the migration's plan sees a real change and recreates. Volumes

@@ -58,17 +58,17 @@ Resources: 0 created, 1 changed, 0 deleted, 0 unchanged
   "schema_name": "default"
 }
 
-=== update the name, this should recreate
+=== update the name, this renames the model in place
 >>> [CLI] bundle plan
-recreate registered_models.my_registered_model
+update registered_models.my_registered_model
 
-Plan: 1 to add, 0 to change, 1 to delete, 0 unchanged
+Plan: 0 to add, 1 to change, 0 to delete, 0 unchanged
 
 >>> [CLI] bundle deploy
 Uploading bundle files to /Workspace/Users/[USERNAME]/.bundle/deploy-registered-models-basic-[UNIQUE_NAME]/default/files...
-Recreated registered_models.my_registered_model
+Updated registered_models.my_registered_model
 Files: 1 uploaded, 0 deleted
-Resources: 1 created, 0 changed, 1 deleted, 0 unchanged
+Resources: 0 created, 1 changed, 0 deleted, 0 unchanged
 
 >>> [CLI] registered-models get main.default.my-registered-model-updated-[UNIQUE_NAME]
 {

@@ -36,7 +36,7 @@ deploy_registered_model
 export NAME="my-registered-model-updated-$UNIQUE_NAME"
 envsubst < databricks.yml.tmpl > databricks.yml
 
-title "update the name, this should recreate"
+title "update the name, this renames the model in place"
 deploy_registered_model
 
 title "update the catalog name, this should recreate"

@@ -0,0 +1,16 @@
+bundle:
+  name: test-bundle-$UNIQUE_NAME
+
+resources:
+  catalogs:
+    catalog1:
+      name: test_catalog_$UNIQUE_NAME
+  schemas:
+    schema1:
+      catalog_name: ${resources.catalogs.catalog1.name}
+      name: myschema
+  registered_models:
+    model1:
+      catalog_name: ${resources.catalogs.catalog1.name}
+      schema_name: ${resources.schemas.schema1.name}
+      name: mymodel
@@ -0,0 +1,2 @@
+Cloud = true
+EnvMatrix.DMS = ["", "true"]
Original file line number	Diff line number	Diff line change
		@@ -0,0 +1 @@
		* Rename a registered model in place when its `name` changes, keeping its versions, instead of deleting and recreating it. Registered models inside a renamed catalog or schema now follow it to the new name; recreating them failed because they already exist there. ([#7379](https://github.com/databricks/cli/pull/7379))