skypilot-nightly 1.0.0.dev20250413__py3-none-any.whl → 1.0.0.dev20250421__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 = '197c8dd3ea85d23323477e7d7cf69e8dc1b693c6'
+_SKYPILOT_COMMIT_SHA = 'd37ba6f11593127305c73cb1b1a03dc6da8452b7'
 
 
 def _get_git_commit():
@@ -35,7 +35,7 @@ def _get_git_commit():
 
 
 __commit__ = _get_git_commit()
-__version__ = '1.0.0.dev20250413'
+__version__ = '1.0.0.dev20250421'
 __root_dir__ = os.path.dirname(os.path.abspath(__file__))
 
 

sky/adaptors/kubernetes.py

@@ -160,6 +160,13 @@ def api_client(context: Optional[str] = None):
     return kubernetes.client.ApiClient()
 
 
+@_api_logging_decorator('urllib3', logging.ERROR)
+@annotations.lru_cache(scope='request')
+def watch(context: Optional[str] = None):
+    _load_config(context)
+    return kubernetes.watch.Watch()
+
+
 def api_exception():
     return kubernetes.client.rest.ApiException
 

sky/authentication.py

@@ -382,10 +382,10 @@ def setup_kubernetes_authentication(config: Dict[str, Any]) -> Dict[str, Any]:
         network_mode = kubernetes_enums.KubernetesNetworkingMode.from_str(
             network_mode_str)
     except ValueError as e:
-        # Add message saying "Please check: ~/.sky/skyconfig.yaml" to the error
+        # Add message saying "Please check: ~/.sky/config.yaml" to the error
         # message.
         with ux_utils.print_exception_no_traceback():
-            raise ValueError(str(e) + ' Please check: ~/.sky/skyconfig.yaml.') \
+            raise ValueError(str(e) + ' Please check: ~/.sky/config.yaml.') \
                 from None
     _, public_key_path = get_or_generate_keys()
 

sky/backends/backend_utils.py

@@ -278,11 +278,11 @@ def _optimize_file_mounts(yaml_path: str) -> None:
             # the dst.
             mkdir_parent = f'mkdir -p {dst}'
             src_basename = f'{src_basename}/*'
-        mv = (f'cp -r {_REMOTE_RUNTIME_FILES_DIR}/{src_basename} '
+        mv = (f'cp -rf {_REMOTE_RUNTIME_FILES_DIR}/{src_basename} '
               f'{dst_parent_dir}/{dst_basename}')
         fragment = f'({mkdir_parent} && {mv})'
         commands.append(fragment)
-    postprocess_runtime_files_command = ' && '.join(commands)
+    postprocess_runtime_files_command = '; '.join(commands)
 
     setup_commands = yaml_config.get('setup_commands', [])
     if setup_commands:
@@ -682,7 +682,7 @@ def write_cluster_config(
             ssh_proxy_command = ssh_proxy_command_config[region_name]
     logger.debug(f'Using ssh_proxy_command: {ssh_proxy_command!r}')
 
-    # User-supplied global instance tags from ~/.sky/skyconfig.yaml.
+    # User-supplied global instance tags from ~/.sky/config.yaml.
     labels = skypilot_config.get_nested((str(cloud).lower(), 'labels'), {})
     # labels is a dict, which is guaranteed by the type check in
     # schemas.py
@@ -2582,11 +2582,36 @@ def get_clusters(
             logger.info(f'Cluster(s) not found: {bright}{clusters_str}{reset}.')
         records = new_records
 
+    def _update_record_with_resources(record: Optional[Dict[str, Any]]) -> None:
+        """Add the resources to the record."""
+        if record is None:
+            return
+        handle = record['handle']
+        if handle is None:
+            return
+        record['nodes'] = handle.launched_nodes
+        if handle.launched_resources is None:
+            return
+        record['cloud'] = (f'{handle.launched_resources.cloud}'
+                           if handle.launched_resources.cloud else None)
+        record['region'] = (f'{handle.launched_resources.region}'
+                            if handle.launched_resources.region else None)
+        record['cpus'] = (f'{handle.launched_resources.cpus}'
+                          if handle.launched_resources.cpus else None)
+        record['memory'] = (f'{handle.launched_resources.memory}'
+                            if handle.launched_resources.memory else None)
+        record['accelerators'] = (f'{handle.launched_resources.accelerators}'
+                                  if handle.launched_resources.accelerators else
+                                  None)
+
     # Add auth_config to the records
     for record in records:
         _update_record_with_credentials_and_resources_str(record)
 
     if refresh == common.StatusRefreshMode.NONE:
+        # Add resources to the records
+        for record in records:
+            _update_record_with_resources(record)
         return records
 
     plural = 's' if len(records) > 1 else ''
@@ -2662,6 +2687,9 @@ def get_clusters(
         for cluster_name, e in failed_clusters:
             logger.warning(f'  {bright}{cluster_name}{reset}: {e}')
 
+    # Add resources to the records
+    for record in kept_records:
+        _update_record_with_resources(record)
     return kept_records
 
 

sky/backends/cloud_vm_ray_backend.py

@@ -35,6 +35,7 @@ from sky import optimizer
 from sky import provision as provision_lib
 from sky import resources as resources_lib
 from sky import sky_logging
+from sky import skypilot_config
 from sky import task as task_lib
 from sky.backends import backend_utils
 from sky.backends import wheel_utils
@@ -439,18 +440,7 @@ class RayCodeGen:
                 pg = ray_util.placement_group({json.dumps(bundles)}, 'STRICT_SPREAD')
                 plural = 's' if {num_nodes} > 1 else ''
                 node_str = f'{num_nodes} node{{plural}}'
-
-                # We have this `INFO: Tip:` message only for backward
-                # compatibility, because if a cluster has the old SkyPilot version,
-                # it relies on this message to start log streaming.
-                # This message will be skipped for new clusters, because we use
-                # start_streaming_at for the `Waiting for task resources on`
-                # message.
-                # TODO: Remove this message in v0.9.0.
-                message = ('{ux_utils.INDENT_SYMBOL}{colorama.Style.DIM}INFO: '
-                           'Tip: use Ctrl-C to exit log streaming, not kill '
-                           'the job.{colorama.Style.RESET_ALL}\\n')
-                message += ('{ux_utils.INDENT_SYMBOL}{colorama.Style.DIM}'
+                message = ('{ux_utils.INDENT_SYMBOL}{colorama.Style.DIM}'
                             'Waiting for task resources on '
                            f'{{node_str}}.{colorama.Style.RESET_ALL}')
                 print(message, flush=True)
@@ -608,9 +598,6 @@ class RayCodeGen:
             textwrap.dedent(f"""\
             sky_env_vars_dict = {{}}
             sky_env_vars_dict['{constants.SKYPILOT_NODE_IPS}'] = job_ip_list_str
-            # Backward compatibility: Environment starting with `SKY_` is
-            # deprecated. Remove it in v0.9.0.
-            sky_env_vars_dict['SKY_NODE_IPS'] = job_ip_list_str
             sky_env_vars_dict['{constants.SKYPILOT_NUM_NODES}'] = len(job_ip_rank_list)
             """)
         ]
@@ -659,9 +646,6 @@ class RayCodeGen:
         if script is not None:
             script += rclone_flush_script
             sky_env_vars_dict['{constants.SKYPILOT_NUM_GPUS_PER_NODE}'] = {int(math.ceil(num_gpus))!r}
-            # Backward compatibility: Environment starting with `SKY_` is
-            # deprecated. Remove it in v0.9.0.
-            sky_env_vars_dict['SKY_NUM_GPUS_PER_NODE'] = {int(math.ceil(num_gpus))!r}
 
             ip = gang_scheduling_id_to_ip[{gang_scheduling_id!r}]
             rank = job_ip_rank_map[ip]
@@ -678,14 +662,8 @@ class RayCodeGen:
                 name_str = f'{{node_name}}, rank={{rank}},'
                 log_path = os.path.expanduser(os.path.join({log_dir!r}, f'{{rank}}-{{node_name}}.log'))
             sky_env_vars_dict['{constants.SKYPILOT_NODE_RANK}'] = rank
-            # Backward compatibility: Environment starting with `SKY_` is
-            # deprecated. Remove it in v0.9.0.
-            sky_env_vars_dict['SKY_NODE_RANK'] = rank
 
             sky_env_vars_dict['SKYPILOT_INTERNAL_JOB_ID'] = {self.job_id}
-            # Backward compatibility: Environment starting with `SKY_` is
-            # deprecated. Remove it in v0.9.0.
-            sky_env_vars_dict['SKY_INTERNAL_JOB_ID'] = {self.job_id}
 
             futures.append(run_bash_command_with_log \\
                     .options(name=name_str, {options_str}) \\
@@ -1473,7 +1451,7 @@ class RetryingVmProvisioner(object):
                     f'invalid cloud credentials: '
                     f'{common_utils.format_exception(e)}')
             except exceptions.InvalidCloudConfigs as e:
-                # Failed due to invalid user configs in ~/.sky/skyconfig.yaml.
+                # Failed due to invalid user configs in ~/.sky/config.yaml.
                 logger.warning(f'{common_utils.format_exception(e)}')
                 # We should block the entire cloud if the user config is
                 # invalid.
@@ -2065,9 +2043,9 @@ class RetryingVmProvisioner(object):
                                (clouds.Kubernetes, clouds.RunPod)) and
                         controller_utils.Controllers.from_name(cluster_name)
                         is not None):
-                    assert (clouds.CloudImplementationFeatures.AUTOSTOP
-                            in requested_features), requested_features
-                    requested_features.remove(
+                    # If autostop is disabled in config, the feature may not be
+                    # requested, so use discard() instead of remove().
+                    requested_features.discard(
                         clouds.CloudImplementationFeatures.AUTOSTOP)
 
                 # Skip if to_provision.cloud does not support requested features
@@ -4470,7 +4448,9 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
                            (clouds.Kubernetes, clouds.RunPod)) and not down and
                     idle_minutes_to_autostop >= 0):
                 # We should hit this code path only for the controllers on
-                # Kubernetes and RunPod clusters.
+                # Kubernetes and RunPod clusters, because autostop() will
+                # skip the supported feature check. Non-controller k8s/runpod
+                # clusters will have already errored out.
                 controller = controller_utils.Controllers.from_name(
                     handle.cluster_name)
                 assert (controller is not None), handle.cluster_name
@@ -4481,6 +4461,19 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
                     # For SkyServe controllers on Kubernetes: override autostop
                     # behavior to force autodown (instead of no-op)
                     # to avoid dangling controllers.
+
+                    # down = False is the default, but warn the user in case
+                    # they have explicitly specified it.
+                    config_override_down = skypilot_config.get_nested(
+                        (controller.value.controller_type, 'controller',
+                         'autostop', 'down'), None)
+                    if config_override_down is False:  # will not match None
+                        logger.warning(
+                            'SkyServe controller autodown is disabled in the '
+                            '~/.sky/config.yaml configuration file '
+                            '(serve.controller.autostop.down_when_idle), but '
+                            'it is force enabled for Kubernetes clusters.')
+
                     down = True
                 else:
                     logger.info('Auto-stop is not supported for Kubernetes '

sky/backends/wheel_utils.py

@@ -85,6 +85,15 @@ def _build_sky_wheel() -> pathlib.Path:
         for f in setup_files_dir.iterdir():
             if f.is_file() and f.name != 'setup.py':
                 shutil.copy(str(f), str(tmp_dir))
+                if f.name == 'MANIFEST.in':
+                    # Remove the line `sky/dashboard/out`, so we do not
+                    # include the dashboard files in the internal wheel
+                    import fileinput  # pylint: disable=import-outside-toplevel
+                    with fileinput.input(tmp_dir / f.name,
+                                         inplace=True) as file:
+                        for line in file:
+                            if 'sky/dashboard/out' not in line:
+                                print(line, end='')
 
         init_file_path = SKY_PACKAGE_PATH / '__init__.py'
         init_file_content = init_file_path.read_text()

sky/check.py

@@ -142,7 +142,7 @@ def check_capabilities(
     if disallowed_cloud_names:
         disallowed_clouds_hint = (
             '\nNote: The following clouds were disabled because they were not '
-            'included in allowed_clouds in ~/.sky/skyconfig.yaml: '
+            'included in allowed_clouds in ~/.sky/config.yaml: '
             f'{", ".join([c for c in disallowed_cloud_names])}')
     if not all_enabled_clouds:
         echo(