diff --git a/sky/backends/cloud_vm_ray_backend.py b/sky/backends/cloud_vm_ray_backend.py
index b92c05511a4..b2f1bcae629 100644
--- a/sky/backends/cloud_vm_ray_backend.py
+++ b/sky/backends/cloud_vm_ray_backend.py
@@ -3680,7 +3680,8 @@ def run_on_head(
 
     @timeline.event
     def _check_existing_cluster(
-            self, task: task_lib.Task, to_provision: resources_lib.Resources,
+            self, task: task_lib.Task,
+            to_provision: Optional[resources_lib.Resources],
             cluster_name: str) -> RetryingVmProvisioner.ToProvisionConfig:
         """Checks if the cluster exists and returns the provision config.
 
@@ -3690,6 +3691,8 @@ def _check_existing_cluster(
             exceptions.InvalidClusterNameError: If the cluster name is invalid.
             # TODO(zhwu): complete the list of exceptions.
         """
+        handle_before_refresh = global_user_state.get_handle_from_cluster_name(
+            cluster_name)
         prev_cluster_status, handle = (
             backend_utils.refresh_cluster_status_handle(
                 cluster_name, acquire_per_cluster_status_lock=False))
@@ -3713,6 +3716,22 @@ def _check_existing_cluster(
                       if resources.cloud is not None else clouds.Cloud)
         task_cloud.check_cluster_name_is_valid(cluster_name)
 
+        if to_provision is None:
+            logger.info(
+                f'The cluster {cluster_name!r} was autodowned or manually '
+                'terminated on the cloud console. Using the same resources '
+                'as the previously terminated one to provision a new cluster.')
+            # The cluster is recently terminated either by autostop or manually
+            # terminated on the cloud. We should use the previously terminated
+            # resources to provision the cluster.
+            assert isinstance(
+                handle_before_refresh, CloudVmRayResourceHandle), (
+                    f'Trying to launch cluster {cluster_name!r} recently '
+                    'terminated  on the cloud, but the handle is not a '
+                    f'CloudVmRayResourceHandle ({handle_before_refresh}).')
+            to_provision = handle_before_refresh.launched_resources
+            self.check_resources_fit_cluster(handle_before_refresh, task)
+
         cloud = to_provision.cloud
         if isinstance(cloud, clouds.Local):
             # The field ssh_user is specified in the cluster config file.
diff --git a/sky/skylet/providers/oci/query_helper.py b/sky/skylet/providers/oci/query_helper.py
index 28aadf90341..29601192e5d 100644
--- a/sky/skylet/providers/oci/query_helper.py
+++ b/sky/skylet/providers/oci/query_helper.py
@@ -57,7 +57,7 @@ def query_instances_by_tags(cls, tag_filters, region):
 
     @classmethod
     def terminate_instances_by_tags(cls, tag_filters, region) -> int:
-        logger.info(f"Terminate instance by tags: {tag_filters}")
+        logger.debug(f"Terminate instance by tags: {tag_filters}")
         insts = cls.query_instances_by_tags(tag_filters, region)
         fail_count = 0
         for inst in insts:
@@ -73,7 +73,7 @@ def terminate_instances_by_tags(cls, tag_filters, region) -> int:
                 traceback.print_exc()
 
         if fail_count == 0:
-            logger.info(f"Instance teardown result: OK")
+            logger.debug(f"Instance teardown result: OK")
         else:
             logger.warn(f"Instance teardown result: {fail_count} failed!")