skypilot-nightly 1.0.0.dev20250807__py3-none-any.whl → 1.0.0.dev20250808__py3-none-any.whl

sky/__init__.py

@@ -5,7 +5,7 @@ from typing import Optional
 import urllib.request
 
 # Replaced with the current commit when building the wheels.
-_SKYPILOT_COMMIT_SHA = 'a167cba8230b0ffda6baa0c825fa0eb5d5ab4aa4'
+_SKYPILOT_COMMIT_SHA = 'eb83a691489c0c37aae9c22f607469ff78a74e34'
 
 
 def _get_git_commit():
@@ -35,7 +35,7 @@ def _get_git_commit():
 
 
 __commit__ = _get_git_commit()
-__version__ = '1.0.0.dev20250807'
+__version__ = '1.0.0.dev20250808'
 __root_dir__ = os.path.dirname(os.path.abspath(__file__))
 
 

sky/backends/backend_utils.py

@@ -121,6 +121,7 @@ CLUSTER_STATUS_LOCK_TIMEOUT_SECONDS = 20
 _CLUSTER_STATUS_CACHE_DURATION_SECONDS = 2
 
 CLUSTER_FILE_MOUNTS_LOCK_TIMEOUT_SECONDS = 10
+WORKSPACE_LOCK_TIMEOUT_SECONDS = 10
 
 # Remote dir that holds our runtime files.
 _REMOTE_RUNTIME_FILES_DIR = '~/.sky/.runtime_files'
@@ -2760,6 +2761,9 @@ def get_clusters(
     refresh: common.StatusRefreshMode,
     cluster_names: Optional[Union[str, List[str]]] = None,
     all_users: bool = True,
+    # Internal only:
+    # pylint: disable=invalid-name
+    _include_is_managed: bool = False,
 ) -> List[Dict[str, Any]]:
     """Returns a list of cached or optionally refreshed cluster records.
 
@@ -2780,6 +2784,8 @@ def get_clusters(
             names.
         all_users: If True, return clusters from all users. If False, only
             return clusters from the current user.
+        _include_is_managed: Whether to force include clusters created by the
+            controller.
 
     Returns:
         A list of cluster records. If the cluster does not exist or has been
@@ -2788,6 +2794,13 @@ def get_clusters(
     records = global_user_state.get_clusters()
     current_user = common_utils.get_current_user()
 
+    # Filter out clusters created by the controller.
+    if (not env_options.Options.SHOW_DEBUG_INFO.get() and
+            not _include_is_managed):
+        records = [
+            record for record in records if not record.get('is_managed', False)
+        ]
+
     # Filter by user if requested
     if not all_users:
         records = [
@@ -3221,7 +3234,8 @@ def get_endpoints(cluster: str,
             with ux_utils.print_exception_no_traceback():
                 raise ValueError(f'Invalid endpoint {port!r}.') from None
     cluster_records = get_clusters(refresh=common.StatusRefreshMode.NONE,
-                                   cluster_names=[cluster])
+                                   cluster_names=[cluster],
+                                   _include_is_managed=True)
     if not cluster_records:
         with ux_utils.print_exception_no_traceback():
             raise exceptions.ClusterNotUpError(
@@ -3311,3 +3325,8 @@ def cluster_status_lock_id(cluster_name: str) -> str:
 def cluster_file_mounts_lock_id(cluster_name: str) -> str:
     """Get the lock ID for cluster file mounts operations."""
     return f'{cluster_name}_file_mounts'
+
+
+def workspace_lock_id(workspace_name: str) -> str:
+    """Get the lock ID for workspace operations."""
+    return f'{workspace_name}_workspace'

sky/backends/cloud_vm_ray_backend.py

@@ -1177,7 +1177,8 @@ class RetryingVmProvisioner(object):
                  local_wheel_path: pathlib.Path,
                  wheel_hash: str,
                  blocked_resources: Optional[Iterable[
-                     resources_lib.Resources]] = None):
+                     resources_lib.Resources]] = None,
+                 is_managed: Optional[bool] = None):
         self._blocked_resources: Set[resources_lib.Resources] = set()
         if blocked_resources:
             # blocked_resources is not None and not empty.
@@ -1189,6 +1190,7 @@ class RetryingVmProvisioner(object):
         self._requested_features = requested_features
         self._local_wheel_path = local_wheel_path
         self._wheel_hash = wheel_hash
+        self._is_managed = is_managed
 
     def _yield_zones(
             self, to_provision: resources_lib.Resources, num_nodes: int,
@@ -1522,6 +1524,7 @@ class RetryingVmProvisioner(object):
                 cluster_handle=handle,
                 requested_resources=requested_resources,
                 ready=False,
+                is_managed=self._is_managed,
             )
 
             global_user_state.set_owner_identity_for_cluster(
@@ -2753,6 +2756,8 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
         self._dag = None
         self._optimize_target = None
         self._requested_features = set()
+        self._dump_final_script = False
+        self._is_managed = False
 
         # Command for running the setup script. It is only set when the
         # setup needs to be run outside the self._setup() and as part of
@@ -2769,6 +2774,7 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
         self._requested_features = kwargs.pop('requested_features',
                                               self._requested_features)
         self._dump_final_script = kwargs.pop('dump_final_script', False)
+        self._is_managed = kwargs.pop('is_managed', False)
         assert not kwargs, f'Unexpected kwargs: {kwargs}'
 
     def check_resources_fit_cluster(
@@ -2990,7 +2996,8 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
                         self._requested_features,
                         local_wheel_path,
                         wheel_hash,
-                        blocked_resources=task.blocked_resources)
+                        blocked_resources=task.blocked_resources,
+                        is_managed=self._is_managed)
                     log_path = os.path.join(self.log_dir, 'provision.log')
                     rich_utils.force_update_status(
                         ux_utils.spinner_message('Launching', log_path))

sky/client/cli/command.py

@@ -35,7 +35,7 @@ import sys
 import traceback
 import typing
 from typing import (Any, Callable, Dict, Generator, List, Optional, Set, Tuple,
-                    Union)
+                    TypeVar, Union)
 
 import click
 import colorama
@@ -116,6 +116,8 @@ _DAG_NOT_SUPPORTED_MESSAGE = ('YAML specifies a DAG which is only supported by '
                               '`sky jobs launch`. `{command}` supports a '
                               'single task only.')
 
+T = TypeVar('T')
+
 
 def _get_cluster_records_and_set_ssh_config(
     clusters: Optional[List[str]],
@@ -224,8 +226,8 @@ def _get_glob_matches(candidate_names: List[str],
     return list(set(glob_storages))
 
 
-def _async_call_or_wait(request_id: str, async_call: bool,
-                        request_name: str) -> Any:
+def _async_call_or_wait(request_id: server_common.RequestId[T],
+                        async_call: bool, request_name: str) -> Any:
     short_request_id = request_id[:8]
     if not async_call:
         try:
@@ -1411,7 +1413,7 @@ def exec(
 
 
 def _handle_jobs_queue_request(
-        request_id: str,
+        request_id: server_common.RequestId[List[Dict[str, Any]]],
         show_all: bool,
         show_user: bool,
         max_num_jobs_to_show: Optional[int],
@@ -1492,7 +1494,7 @@ def _handle_jobs_queue_request(
 
 
 def _handle_services_request(
-    request_id: str,
+    request_id: server_common.RequestId[List[Dict[str, Any]]],
     service_names: Optional[List[str]],
     show_all: bool,
     show_endpoint: bool,
@@ -1879,17 +1881,19 @@ def status(verbose: bool, refresh: bool, ip: bool, endpoints: bool,
                                   skip_finished=True,
                                   all_users=all_users)
 
-    def submit_services() -> Optional[str]:
+    def submit_services(
+    ) -> Optional[server_common.RequestId[List[Dict[str, Any]]]]:
         return serve_lib.status(service_names=None)
 
-    def submit_pools() -> Optional[str]:
+    def submit_pools(
+    ) -> Optional[server_common.RequestId[List[Dict[str, Any]]]]:
         try:
             return managed_jobs.pool_status(pool_names=None)
         except exceptions.APINotSupportedError as e:
             logger.debug(f'Pools are not supported in the remote server: {e}')
             return None
 
-    def submit_workspace() -> Optional[str]:
+    def submit_workspace() -> Optional[server_common.RequestId[Dict[str, Any]]]:
         try:
             return sdk.workspaces()
         except RuntimeError:
@@ -1928,11 +1932,14 @@ def status(verbose: bool, refresh: bool, ip: bool, endpoints: bool,
         if not (ip or show_endpoints):
             workspace_request_id = workspace_request_future.result()
 
-    managed_jobs_queue_request_id = ('' if not managed_jobs_queue_request_id
-                                     else managed_jobs_queue_request_id)
-    service_status_request_id = ('' if not service_status_request_id else
+    managed_jobs_queue_request_id = (server_common.RequestId()
+                                     if not managed_jobs_queue_request_id else
+                                     managed_jobs_queue_request_id)
+    service_status_request_id = (server_common.RequestId()
+                                 if not service_status_request_id else
                                  service_status_request_id)
-    pool_status_request_id = ('' if not pool_status_request_id else
+    pool_status_request_id = (server_common.RequestId()
+                              if not pool_status_request_id else
                               pool_status_request_id)
 
     # Phase 3: Get cluster records and handle special cases
@@ -1957,7 +1964,7 @@ def status(verbose: bool, refresh: bool, ip: bool, endpoints: bool,
     if workspace_request_id is not None:
         all_workspaces = sdk.get(workspace_request_id)
     else:
-        all_workspaces = [constants.SKYPILOT_DEFAULT_WORKSPACE]
+        all_workspaces = {constants.SKYPILOT_DEFAULT_WORKSPACE: {}}
     active_workspace = skypilot_config.get_active_workspace()
     show_workspace = len(all_workspaces) > 1
     _show_enabled_infra(active_workspace, show_workspace)
@@ -3836,7 +3843,7 @@ def show_gpus(
                 yield k8s_messages
                 yield '\n\n'
 
-            result = sdk.stream_and_get(
+            list_accelerator_counts_result = sdk.stream_and_get(
                 sdk.list_accelerator_counts(
                     gpus_only=True,
                     clouds=clouds_to_list,
@@ -3853,14 +3860,20 @@ def show_gpus(
 
             # "Common" GPUs
             for gpu in catalog.get_common_gpus():
-                if gpu in result:
-                    gpu_table.add_row([gpu, _list_to_str(result.pop(gpu))])
+                if gpu in list_accelerator_counts_result:
+                    gpu_table.add_row([
+                        gpu,
+                        _list_to_str(list_accelerator_counts_result.pop(gpu))
+                    ])
             yield from gpu_table.get_string()
 
             # Google TPUs
             for tpu in catalog.get_tpus():
-                if tpu in result:
-                    tpu_table.add_row([tpu, _list_to_str(result.pop(tpu))])
+                if tpu in list_accelerator_counts_result:
+                    tpu_table.add_row([
+                        tpu,
+                        _list_to_str(list_accelerator_counts_result.pop(tpu))
+                    ])
             if tpu_table.get_string():
                 yield '\n\n'
             yield from tpu_table.get_string()
@@ -3868,7 +3881,7 @@ def show_gpus(
             # Other GPUs
             if show_all:
                 yield '\n\n'
-                for gpu, qty in sorted(result.items()):
+                for gpu, qty in sorted(list_accelerator_counts_result.items()):
                     other_table.add_row([gpu, _list_to_str(qty)])
                 yield from other_table.get_string()
                 yield '\n\n'
@@ -3919,7 +3932,7 @@ def show_gpus(
 
         # For clouds other than Kubernetes, get the accelerator details
         # Case-sensitive
-        result = sdk.stream_and_get(
+        list_accelerators_result = sdk.stream_and_get(
             sdk.list_accelerators(gpus_only=True,
                                   name_filter=name,
                                   quantity_filter=quantity,
@@ -3935,8 +3948,8 @@ def show_gpus(
         #   - Group by cloud
         #   - Sort within each group by prices
         #   - Sort groups by each cloud's (min price, min spot price)
-        new_result = {}
-        for i, (gpu, items) in enumerate(result.items()):
+        new_result: Dict[str, List[catalog_common.InstanceTypeInfo]] = {}
+        for i, (gpu, items) in enumerate(list_accelerators_result.items()):
             df = pd.DataFrame([t._asdict() for t in items])
             # Determine the minimum prices for each cloud.
             min_price_df = df.groupby('cloud').agg(min_price=('price', 'min'),
@@ -3954,14 +3967,14 @@ def show_gpus(
                 for row in df.to_records(index=False)
             ]
             new_result[gpu] = sorted_dataclasses
-        result = new_result
+        list_accelerators_result = new_result
 
         if print_section_titles and not show_all:
             yield '\n\n'
             yield (f'{colorama.Fore.CYAN}{colorama.Style.BRIGHT}'
                    f'Cloud GPUs{colorama.Style.RESET_ALL}\n')
 
-        if not result:
+        if not list_accelerators_result:
             quantity_str = (f' with requested quantity {quantity}'
                             if quantity else '')
             cloud_str = f' on {cloud_obj}.' if cloud_name else ' in cloud catalogs.'
@@ -3969,7 +3982,7 @@ def show_gpus(
             yield 'To show available accelerators, run: sky show-gpus --all'
             return
 
-        for i, (gpu, items) in enumerate(result.items()):
+        for i, (gpu, items) in enumerate(list_accelerators_result.items()):
             accelerator_table_headers = [
                 'GPU',
                 'QTY',
@@ -6039,7 +6052,8 @@ def api_logs(request_id: Optional[str], server_logs: bool,
     if request_id is not None and log_path is not None:
         raise click.BadParameter(
             'Only one of request ID and log path can be provided.')
-    sdk.stream_and_get(request_id, log_path, tail)
+    sdk.stream_and_get(server_common.RequestId[None](request_id), log_path,
+                       tail)
 
 
 @api.command('cancel', cls=_DocumentedCodeCommand)