skypilot-nightly 1.0.0.dev20250627__py3-none-any.whl → 1.0.0.dev20250630__py3-none-any.whl

sky/utils/schemas.py

@@ -7,6 +7,7 @@ import enum
 from typing import Any, Dict, List, Tuple
 
 from sky.skylet import constants
+from sky.utils import kubernetes_enums
 
 
 def _check_not_both_fields_present(field1: str, field2: str):
@@ -1018,10 +1019,73 @@ _REMOTE_IDENTITY_SCHEMA_KUBERNETES = {
     },
 }
 
+_CONTEXT_CONFIG_SCHEMA_KUBERNETES = {
+    'networking': {
+        'type': 'string',
+        'case_insensitive_enum': [
+            type.value for type in kubernetes_enums.KubernetesNetworkingMode
+        ],
+    },
+    'ports': {
+        'type': 'string',
+        'case_insensitive_enum': [
+            type.value for type in kubernetes_enums.KubernetesPortMode
+        ],
+    },
+    'pod_config': {
+        'type': 'object',
+        'required': [],
+        # Allow arbitrary keys since validating pod spec is hard
+        'additionalProperties': True,
+    },
+    'custom_metadata': {
+        'type': 'object',
+        'required': [],
+        # Allow arbitrary keys since validating metadata is hard
+        'additionalProperties': True,
+        # Disallow 'name' and 'namespace' keys in this dict
+        'not': {
+            'anyOf': [{
+                'required': ['name']
+            }, {
+                'required': ['namespace']
+            }]
+        },
+    },
+    'provision_timeout': {
+        'type': 'integer',
+    },
+    'autoscaler': {
+        'type': 'string',
+        'case_insensitive_enum': [
+            type.value for type in kubernetes_enums.KubernetesAutoscalerType
+        ],
+    },
+    'high_availability': {
+        'type': 'object',
+        'required': [],
+        'additionalProperties': False,
+        'properties': {
+            'storage_class_name': {
+                'type': 'string',
+            }
+        },
+    },
+    'kueue': {
+        'type': 'object',
+        'required': [],
+        'additionalProperties': False,
+        'properties': {
+            'local_queue_name': {
+                'type': 'string',
+            },
+        },
+    },
+}
+
 
 def get_config_schema():
     # pylint: disable=import-outside-toplevel
-    from sky.utils import kubernetes_enums
 
     resources_schema = {
         k: v
@@ -1178,70 +1242,21 @@ def get_config_schema():
                         'type': 'string',
                     },
                 },
-                'networking': {
-                    'type': 'string',
-                    'case_insensitive_enum': [
-                        type.value
-                        for type in kubernetes_enums.KubernetesNetworkingMode
-                    ]
-                },
-                'ports': {
-                    'type': 'string',
-                    'case_insensitive_enum': [
-                        type.value
-                        for type in kubernetes_enums.KubernetesPortMode
-                    ]
-                },
-                'pod_config': {
+                'context_configs': {
                     'type': 'object',
                     'required': [],
-                    # Allow arbitrary keys since validating pod spec is hard
-                    'additionalProperties': True,
-                },
-                'custom_metadata': {
-                    'type': 'object',
-                    'required': [],
-                    # Allow arbitrary keys since validating metadata is hard
-                    'additionalProperties': True,
-                    # Disallow 'name' and 'namespace' keys in this dict
-                    'not': {
-                        'anyOf': [{
-                            'required': ['name']
-                        }, {
-                            'required': ['namespace']
-                        }]
-                    }
-                },
-                'provision_timeout': {
-                    'type': 'integer',
-                },
-                'autoscaler': {
-                    'type': 'string',
-                    'case_insensitive_enum': [
-                        type.value
-                        for type in kubernetes_enums.KubernetesAutoscalerType
-                    ]
-                },
-                'high_availability': {
-                    'type': 'object',
-                    'required': [],
-                    'additionalProperties': False,
-                    'properties': {
-                        'storage_class_name': {
-                            'type': 'string',
-                        }
-                    }
-                },
-                'kueue': {
-                    'type': 'object',
-                    'required': [],
-                    'additionalProperties': False,
-                    'properties': {
-                        'local_queue_name': {
-                            'type': 'string',
+                    'properties': {},
+                    # Properties are kubernetes context names.
+                    'additionalProperties': {
+                        'type': 'object',
+                        'required': [],
+                        'additionalProperties': False,
+                        'properties': {
+                            **_CONTEXT_CONFIG_SCHEMA_KUBERNETES,
                         },
                     },
                 },
+                **_CONTEXT_CONFIG_SCHEMA_KUBERNETES,
             }
         },
         'ssh': {
@@ -1400,6 +1415,18 @@ def get_config_schema():
                 # Apply validation for URL
                 'pattern': r'^https?://.*$',
             },
+            'service_account_token': {
+                'anyOf': [
+                    {
+                        'type': 'string',
+                        # Validate that token starts with sky_ prefix
+                        'pattern': r'^sky_.+$',
+                    },
+                    {
+                        'type': 'null',
+                    }
+                ]
+            },
         }
     }
 

sky/utils/subprocess_utils.py

@@ -6,6 +6,7 @@ import random
 import resource
 import shlex
 import subprocess
+import sys
 import threading
 import time
 import typing
@@ -16,7 +17,6 @@ import colorama
 from sky import exceptions
 from sky import sky_logging
 from sky.adaptors import common as adaptors_common
-from sky.skylet import constants
 from sky.skylet import log_lib
 from sky.utils import common_utils
 from sky.utils import timeline
@@ -322,12 +322,8 @@ def kill_process_daemon(process_pid: int) -> None:
     daemon_script = os.path.join(
         os.path.dirname(os.path.abspath(log_lib.__file__)),
         'subprocess_daemon.py')
-    python_path = subprocess.check_output(constants.SKY_GET_PYTHON_PATH_CMD,
-                                          shell=True,
-                                          stderr=subprocess.DEVNULL,
-                                          encoding='utf-8').strip()
     daemon_cmd = [
-        python_path,
+        sys.executable,
         daemon_script,
         '--parent-pid',
         str(parent_pid),

sky/volumes/server/core.py

@@ -13,7 +13,9 @@ from sky import models
 from sky import provision
 from sky import sky_logging
 from sky.utils import common_utils
+from sky.utils import rich_utils
 from sky.utils import status_lib
+from sky.utils import ux_utils
 
 logger = sky_logging.init_logger(__name__)
 
@@ -33,14 +35,14 @@ def volume_refresh():
                            'Skipping status refresh...')
             continue
         cloud = config.cloud
-        usedby = provision.get_volume_usedby(cloud, config)
+        usedby_pods, _ = provision.get_volume_usedby(cloud, config)
         with _volume_lock(volume_name):
             latest_volume = global_user_state.get_volume_by_name(volume_name)
             if latest_volume is None:
                 logger.warning(f'Volume {volume_name} not found.')
                 continue
             status = latest_volume.get('status')
-            if not usedby:
+            if not usedby_pods:
                 if status != status_lib.VolumeStatus.READY:
                     logger.info(f'Update volume {volume_name} '
                                 f'status to READY')
@@ -49,7 +51,7 @@ def volume_refresh():
             else:
                 if status != status_lib.VolumeStatus.IN_USE:
                     logger.info(f'Update volume {volume_name} '
-                                f'status to IN_USE, usedby: {usedby}')
+                                f'status to IN_USE, usedby: {usedby_pods}')
                     global_user_state.update_volume_status(
                         volume_name, status=status_lib.VolumeStatus.IN_USE)
 
@@ -74,42 +76,52 @@ def volume_list() -> List[Dict[str, Any]]:
                 'last_attached_at': int timestamp of last attachment,
                 'last_use': last command,
                 'status': sky.VolumeStatus,
+                'usedby_pods': List[str],
+                'usedby_clusters': List[str],
             }
         ]
     """
-    volumes = global_user_state.get_volumes()
-    all_users = global_user_state.get_all_users()
-    user_map = {user.id: user.name for user in all_users}
-    records = []
-    for volume in volumes:
-        volume_name = volume.get('name')
-        record = {
-            'name': volume_name,
-            'launched_at': volume.get('launched_at'),
-            'user_hash': volume.get('user_hash'),
-            'user_name': user_map.get(volume.get('user_hash'), ''),
-            'workspace': volume.get('workspace'),
-            'last_attached_at': volume.get('last_attached_at'),
-            'last_use': volume.get('last_use'),
-        }
-        status = volume.get('status')
-        if status is not None:
-            record['status'] = status.value
-        else:
-            record['status'] = ''
-        config = volume.get('handle')
-        if config is None:
-            logger.warning(f'Volume {volume_name} has no handle.')
-            continue
-        record['type'] = config.type
-        record['cloud'] = config.cloud
-        record['region'] = config.region
-        record['zone'] = config.zone
-        record['size'] = config.size
-        record['config'] = config.config
-        record['name_on_cloud'] = config.name_on_cloud
-        records.append(record)
-    return records
+    with rich_utils.safe_status(ux_utils.spinner_message('Listing volumes')):
+        volumes = global_user_state.get_volumes()
+        all_users = global_user_state.get_all_users()
+        user_map = {user.id: user.name for user in all_users}
+        records = []
+        for volume in volumes:
+            volume_name = volume.get('name')
+            record = {
+                'name': volume_name,
+                'launched_at': volume.get('launched_at'),
+                'user_hash': volume.get('user_hash'),
+                'user_name': user_map.get(volume.get('user_hash'), ''),
+                'workspace': volume.get('workspace'),
+                'last_attached_at': volume.get('last_attached_at'),
+                'last_use': volume.get('last_use'),
+                'usedby_pods': [],
+                'usedby_clusters': [],
+            }
+            status = volume.get('status')
+            if status is not None:
+                record['status'] = status.value
+            else:
+                record['status'] = ''
+            config = volume.get('handle')
+            if config is None:
+                logger.warning(f'Volume {volume_name} has no handle.')
+                continue
+            cloud = config.cloud
+            usedby_pods, usedby_clusters = provision.get_volume_usedby(
+                cloud, config)
+            record['type'] = config.type
+            record['cloud'] = config.cloud
+            record['region'] = config.region
+            record['zone'] = config.zone
+            record['size'] = config.size
+            record['config'] = config.config
+            record['name_on_cloud'] = config.name_on_cloud
+            record['usedby_pods'] = usedby_pods
+            record['usedby_clusters'] = usedby_clusters
+            records.append(record)
+        return records
 
 
 def volume_delete(names: List[str]) -> None:
@@ -122,20 +134,32 @@ def volume_delete(names: List[str]) -> None:
         ValueError: If the volume does not exist
           or is in use or has no handle.
     """
-    for name in names:
-        volume = global_user_state.get_volume_by_name(name)
-        if volume is None:
-            raise ValueError(f'Volume {name} not found.')
-        if volume.get('status') == status_lib.VolumeStatus.IN_USE:
-            raise ValueError(f'Volume {name} is in use.')
-        config = volume.get('handle')
-        if config is None:
-            raise ValueError(f'Volume {name} has no handle.')
-        logger.debug(f'Deleting volume {name} with config {config}')
-        cloud = config.cloud
-        with _volume_lock(name):
-            provision.delete_volume(cloud, config)
-            global_user_state.delete_volume(name)
+    with rich_utils.safe_status(ux_utils.spinner_message('Deleting volumes')):
+        for name in names:
+            volume = global_user_state.get_volume_by_name(name)
+            if volume is None:
+                raise ValueError(f'Volume {name} not found.')
+            config = volume.get('handle')
+            if config is None:
+                raise ValueError(f'Volume {name} has no handle.')
+            cloud = config.cloud
+            usedby_pods, usedby_clusters = provision.get_volume_usedby(
+                cloud, config)
+            if usedby_clusters:
+                usedby_clusters_str = ', '.join(usedby_clusters)
+                cluster_str = 'clusters' if len(
+                    usedby_clusters) > 1 else 'cluster'
+                raise ValueError(f'Volume {name} is used by {cluster_str}'
+                                 f' {usedby_clusters_str}.')
+            if usedby_pods:
+                usedby_pods_str = ', '.join(usedby_pods)
+                pod_str = 'pods' if len(usedby_pods) > 1 else 'pod'
+                raise ValueError(
+                    f'Volume {name} is used by {pod_str} {usedby_pods_str}.')
+            logger.debug(f'Deleting volume {name} with config {config}')
+            with _volume_lock(name):
+                provision.delete_volume(cloud, config)
+                global_user_state.delete_volume(name)
 
 
 def volume_apply(name: str, volume_type: str, cloud: str, region: Optional[str],
@@ -153,34 +177,35 @@ def volume_apply(name: str, volume_type: str, cloud: str, region: Optional[str],
         config: The configuration of the volume.
 
     """
-    # Reuse the method for cluster name on cloud to
-    # generate the storage name on cloud.
-    cloud_obj = sky.CLOUD_REGISTRY.from_str(cloud)
-    assert cloud_obj is not None
-    name_uuid = str(uuid.uuid4())[:6]
-    name_on_cloud = common_utils.make_cluster_name_on_cloud(
-        name, max_length=cloud_obj.max_cluster_name_length())
-    name_on_cloud += '-' + name_uuid
-    config = models.VolumeConfig(
-        name=name,
-        type=volume_type,
-        cloud=str(cloud_obj),
-        region=region,
-        zone=zone,
-        size=size,
-        config=config,
-        name_on_cloud=name_on_cloud,
-    )
-    logger.debug(
-        f'Creating volume {name} on cloud {cloud} with config {config}')
-    with _volume_lock(name):
-        current_volume = global_user_state.get_volume_by_name(name)
-        if current_volume is not None:
-            logger.info(f'Volume {name} already exists.')
-            return
-        config = provision.apply_volume(cloud, config)
-        global_user_state.add_volume(name, config,
-                                     status_lib.VolumeStatus.READY)
+    with rich_utils.safe_status(ux_utils.spinner_message('Creating volume')):
+        # Reuse the method for cluster name on cloud to
+        # generate the storage name on cloud.
+        cloud_obj = sky.CLOUD_REGISTRY.from_str(cloud)
+        assert cloud_obj is not None
+        name_uuid = str(uuid.uuid4())[:6]
+        name_on_cloud = common_utils.make_cluster_name_on_cloud(
+            name, max_length=cloud_obj.max_cluster_name_length())
+        name_on_cloud += '-' + name_uuid
+        config = models.VolumeConfig(
+            name=name,
+            type=volume_type,
+            cloud=str(cloud_obj),
+            region=region,
+            zone=zone,
+            size=size,
+            config=config,
+            name_on_cloud=name_on_cloud,
+        )
+        logger.debug(
+            f'Creating volume {name} on cloud {cloud} with config {config}')
+        with _volume_lock(name):
+            current_volume = global_user_state.get_volume_by_name(name)
+            if current_volume is not None:
+                logger.info(f'Volume {name} already exists.')
+                return
+            config = provision.apply_volume(cloud, config)
+            global_user_state.add_volume(name, config,
+                                         status_lib.VolumeStatus.READY)
 
 
 @contextlib.contextmanager

sky/volumes/utils.py

@@ -6,6 +6,8 @@ from typing import Any, Dict, List, Optional
 import prettytable
 
 from sky import sky_logging
+from sky.skylet import constants
+from sky.utils import common_utils
 from sky.utils import log_utils
 from sky.volumes import volume
 
@@ -44,12 +46,12 @@ class PVCVolumeTable(VolumeTable):
 
     def _create_table(self, show_all: bool = False) -> prettytable.PrettyTable:
         """Create the PVC volume table."""
-        #  If show_all is True, show the table with the columns:
-        #   NAME, TYPE, INFRA, SIZE, USER, WORKSPACE,
-        #   AGE, LAST_USE, STATUS
         #  If show_all is False, show the table with the columns:
         #   NAME, TYPE, INFRA, SIZE, USER, WORKSPACE,
-        #   AGE, LAST_USE, STATUS, NAME_ON_CLOUD,
+        #   AGE, STATUS, LAST_USE, USED_BY
+        #  If show_all is True, show the table with the columns:
+        #   NAME, TYPE, INFRA, SIZE, USER, WORKSPACE,
+        #   AGE, STATUS, LAST_USE, USED_BY, NAME_ON_CLOUD
         #   STORAGE_CLASS, ACCESS_MODE
 
         if show_all:
@@ -63,6 +65,7 @@ class PVCVolumeTable(VolumeTable):
                 'AGE',
                 'STATUS',
                 'LAST_USE',
+                'USED_BY',
                 'NAME_ON_CLOUD',
                 'STORAGE_CLASS',
                 'ACCESS_MODE',
@@ -78,6 +81,7 @@ class PVCVolumeTable(VolumeTable):
                 'AGE',
                 'STATUS',
                 'LAST_USE',
+                'USED_BY',
             ]
 
         table = log_utils.create_table(columns)
@@ -98,6 +102,18 @@ class PVCVolumeTable(VolumeTable):
             size = row.get('size', '')
             if size:
                 size = f'{size}Gi'
+            usedby_str = '-'
+            usedby_clusters = row.get('usedby_clusters')
+            usedby_pods = row.get('usedby_pods')
+            if usedby_clusters:
+                usedby_str = f'{", ".join(usedby_clusters)}'
+            elif usedby_pods:
+                usedby_str = f'{", ".join(usedby_pods)}'
+            if show_all:
+                usedby = usedby_str
+            else:
+                usedby = common_utils.truncate_long_string(
+                    usedby_str, constants.USED_BY_TRUNC_LENGTH)
             infra = _get_infra_str(row.get('cloud'), row.get('region'),
                                    row.get('zone'))
             table_row = [
@@ -107,9 +123,10 @@ class PVCVolumeTable(VolumeTable):
                 size,
                 row.get('user_name', '-'),
                 row.get('workspace', '-'),
-                log_utils.readable_time_duration(row.get('launched_at', 0)),
+                log_utils.human_duration(row.get('launched_at', 0)),
                 row.get('status', ''),
                 last_attached_at_str,
+                usedby,
             ]
             if show_all:
                 table_row.append(row.get('name_on_cloud', ''))