skypilot-nightly 1.0.0.dev20250707__tar.gz → 1.0.0.dev20250709__tar.gz

{skypilot_nightly-1.0.0.dev20250707/skypilot_nightly.egg-info → skypilot_nightly-1.0.0.dev20250709}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: skypilot-nightly
-Version: 1.0.0.dev20250707
+Version: 1.0.0.dev20250709
 Summary: SkyPilot: Run AI on Any Infra — Unified, Faster, Cheaper.
 Author: SkyPilot Team
 License: Apache 2.0

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/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 = '37ee86178d5bc23c6dc4fd69f0294b15075a64b8'
+_SKYPILOT_COMMIT_SHA = '049a61037fec870c541621cb37b03783e046d44b'
 
 
 def _get_git_commit():
@@ -35,7 +35,7 @@ def _get_git_commit():
 
 
 __commit__ = _get_git_commit()
-__version__ = '1.0.0.dev20250707'
+__version__ = '1.0.0.dev20250709'
 __root_dir__ = os.path.dirname(os.path.abspath(__file__))
 
 

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/backends/cloud_vm_ray_backend.py

@@ -1600,7 +1600,8 @@ class RetryingVmProvisioner(object):
                     CloudVmRayBackend().post_teardown_cleanup(
                         handle,
                         terminate=not prev_cluster_ever_up,
-                        remove_from_db=False)
+                        remove_from_db=False,
+                        failover=True)
                     # TODO(suquark): other clouds may have different zone
                     #  blocking strategy. See '_update_blocklist_on_error'
                     #  for details.
@@ -4402,12 +4403,17 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
                               handle: CloudVmRayResourceHandle,
                               terminate: bool,
                               purge: bool = False,
-                              remove_from_db: bool = True) -> None:
+                              remove_from_db: bool = True,
+                              failover: bool = False) -> None:
         """Cleanup local configs/caches and delete TPUs after teardown.
 
         This method will handle the following cleanup steps:
         * Deleting the TPUs;
         * Removing ssh configs for the cluster;
+        * Deleting the open ports;
+        * Deleting the custom multi network infrastructure based on the
+          failover flag (e.g. delete firewalls, subnets, and VPCs for GPU
+          Direct if failover is False, otherwise, only delete the subnets);
         * Updating the local state of the cluster;
         * Removing the terminated cluster's scripts and ray yaml files.
         """
@@ -4439,12 +4445,14 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
             # The cluster yaml does not exist when skypilot has not found
             # the right resource to provision the cluster.
             if handle.cluster_yaml is not None:
+                launched_resources = (
+                    handle.launched_resources.assert_launchable())
+                cloud = launched_resources.cloud
+                config = global_user_state.get_cluster_yaml_dict(
+                    handle.cluster_yaml)
+                ports_cleaned_up = False
+                custom_multi_network_cleaned_up = False
                 try:
-                    launched_resources = (
-                        handle.launched_resources.assert_launchable())
-                    cloud = launched_resources.cloud
-                    config = global_user_state.get_cluster_yaml_dict(
-                        handle.cluster_yaml)
                     cloud.check_features_are_supported(
                         launched_resources,
                         {clouds.CloudImplementationFeatures.OPEN_PORTS})
@@ -4452,7 +4460,7 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
                                                 cluster_name_on_cloud,
                                                 handle.launched_resources.ports,
                                                 config['provider'])
-                    self.remove_cluster_config(handle)
+                    ports_cleaned_up = True
                 except exceptions.NotSupportedError:
                     pass
                 except exceptions.PortDoesNotExistError:
@@ -4466,6 +4474,42 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
                     else:
                         raise
 
+                # Clean up custom multi networks, e.g. the subnets, firewalls,
+                # and VPCs created for GCP GPUDirect TCPX
+                try:
+                    cloud.check_features_are_supported(
+                        handle.launched_resources, {
+                            clouds.CloudImplementationFeatures.
+                            CUSTOM_MULTI_NETWORK
+                        })
+                    provision_lib.cleanup_custom_multi_network(
+                        repr(cloud), cluster_name_on_cloud, config['provider'],
+                        failover)
+                    custom_multi_network_cleaned_up = True
+                except exceptions.NotSupportedError:
+                    pass
+                except Exception as e:  # pylint: disable=broad-except
+                    if purge:
+                        msg = common_utils.format_exception(e, use_bracket=True)
+                        logger.warning(
+                            f'Failed to cleanup custom multi network. Skipping '
+                            f'since purge is set. Details: {msg}')
+                    else:
+                        raise
+
+                if ports_cleaned_up and custom_multi_network_cleaned_up:
+                    try:
+                        self.remove_cluster_config(handle)
+                    except Exception as e:  # pylint: disable=broad-except
+                        if purge:
+                            msg = common_utils.format_exception(
+                                e, use_bracket=True)
+                            logger.warning(
+                                f'Failed to remove cluster config. Skipping '
+                                f'since purge is set. Details: {msg}')
+                        else:
+                            raise
+
         sky.utils.cluster_utils.SSHConfigHelper.remove_cluster(
             handle.cluster_name)
 

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/client/common.py

@@ -206,9 +206,14 @@ def _upload_chunk_with_retry(params: UploadChunkParams) -> None:
                     f'Retrying... ({attempt + 1} / {max_attempts})')
                 time.sleep(1)
             else:
+                try:
+                    response_details = response.json().get('detail')
+                except Exception:  # pylint: disable=broad-except
+                    response_details = response.content
                 error_msg = (
                     f'Failed to upload chunk: {params.chunk_index + 1} / '
-                    f'{params.total_chunks}: {response.json().get("detail")}')
+                    f'{params.total_chunks}: {response_details} '
+                    f'(Status code: {response.status_code})')
                 upload_logger.error(error_msg)
                 with ux_utils.print_exception_no_traceback():
                     raise RuntimeError(

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/clouds/aws.py

@@ -174,6 +174,11 @@ class AWS(clouds.Cloud):
                 f'High availability controllers are not supported on {cls._REPR}.'
             )
 
+        unsupported_features[
+            clouds.CloudImplementationFeatures.CUSTOM_MULTI_NETWORK] = (
+                f'Customized multiple network interfaces are not supported on {cls._REPR}.'
+            )
+
         return unsupported_features
 
     @classmethod

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/clouds/azure.py

@@ -94,6 +94,9 @@ class Azure(clouds.Cloud):
             clouds.CloudImplementationFeatures.HIGH_AVAILABILITY_CONTROLLERS: (
                 f'High availability controllers are not supported on {cls._REPR}.'
             ),
+            clouds.CloudImplementationFeatures.CUSTOM_MULTI_NETWORK: (
+                f'Customized multiple network interfaces are not supported on {cls._REPR}.'
+            ),
         }
         if resources.use_spot:
             features[clouds.CloudImplementationFeatures.STOP] = (

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/clouds/cloud.py

@@ -55,6 +55,9 @@ class CloudImplementationFeatures(enum.Enum):
     AUTO_TERMINATE = 'auto_terminate'  # Pod/VM can stop or down itself
     AUTOSTOP = 'autostop'  # Pod/VM can stop itself
     AUTODOWN = 'autodown'  # Pod/VM can down itself
+    # Pod/VM can have customized multiple network interfaces
+    # e.g. GCP GPUDirect TCPX
+    CUSTOM_MULTI_NETWORK = 'custom_multi_network'
 
 
 # Use str, enum.Enum to allow CloudCapability to be used as a string.

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/clouds/cudo.py

@@ -73,6 +73,9 @@ class Cudo(clouds.Cloud):
         ),
         clouds.CloudImplementationFeatures.HIGH_AVAILABILITY_CONTROLLERS:
             ('High availability controllers are not supported on Cudo.'),
+        clouds.CloudImplementationFeatures.CUSTOM_MULTI_NETWORK:
+            ('Customized multiple network interfaces are not supported on Cudo.'
+            ),
     }
     _MAX_CLUSTER_NAME_LEN_LIMIT = 60
 

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/clouds/do.py

@@ -40,6 +40,9 @@ class DO(clouds.Cloud):
         clouds.CloudImplementationFeatures.HIGH_AVAILABILITY_CONTROLLERS:
             ('High availability controllers are not supported in '
              f'{_REPR}.'),
+        clouds.CloudImplementationFeatures.CUSTOM_MULTI_NETWORK:
+            ('Customized multiple network interfaces are not supported in '
+             f'{_REPR}.'),
     }
     # DO maximum node name length defined as <= 255
     # https://docs.digitalocean.com/reference/api/api-reference/#operation/droplets_create

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/clouds/fluidstack.py

@@ -63,6 +63,9 @@ class Fluidstack(clouds.Cloud):
         clouds.CloudImplementationFeatures.HIGH_AVAILABILITY_CONTROLLERS:
             ('High availability controllers are not supported in '
              f'{_REPR}.'),
+        clouds.CloudImplementationFeatures.CUSTOM_MULTI_NETWORK:
+            ('Customized multiple network interfaces are not supported in '
+             f'{_REPR}.'),
     }
     # Using the latest SkyPilot provisioner API to provision and check status.
     PROVISIONER_VERSION = clouds.ProvisionerVersion.SKYPILOT

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/clouds/gcp.py

@@ -512,8 +512,9 @@ class GCP(clouds.Cloud):
             default_value=False,
             override_configs=resources.cluster_config_overrides)
         resources_vars['enable_gpu_direct'] = enable_gpu_direct
-        network_tier = r.network_tier
-        resources_vars['network_tier'] = network_tier
+        network_tier = (r.network_tier if r.network_tier is not None else
+                        resources_utils.NetworkTier.STANDARD)
+        resources_vars['network_tier'] = network_tier.value
         accelerators = r.accelerators
         if accelerators is not None:
             assert len(accelerators) == 1, r

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/clouds/ibm.py

@@ -53,6 +53,9 @@ class IBM(clouds.Cloud):
                 (f'Opening ports is currently not supported on {cls._REPR}.'),
             clouds.CloudImplementationFeatures.HIGH_AVAILABILITY_CONTROLLERS:
                 ('High availability controllers are not supported on IBM.'),
+            clouds.CloudImplementationFeatures.CUSTOM_MULTI_NETWORK:
+                ('Customized multiple network interfaces are not supported on '
+                 f'{cls._REPR}.'),
         }
         if resources.use_spot:
             features[clouds.CloudImplementationFeatures.STOP] = (

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/clouds/kubernetes.py

@@ -14,6 +14,7 @@ from sky import exceptions
 from sky import sky_logging
 from sky import skypilot_config
 from sky.adaptors import kubernetes
+from sky.clouds.utils import gcp_utils
 from sky.provision import instance_setup
 from sky.provision.kubernetes import network_utils
 from sky.provision.kubernetes import utils as kubernetes_utils
@@ -22,6 +23,7 @@ from sky.provision.kubernetes.utils import normalize_tpu_accelerator_name
 from sky.skylet import constants
 from sky.utils import annotations
 from sky.utils import common_utils
+from sky.utils import kubernetes_enums
 from sky.utils import registry
 from sky.utils import resources_utils
 from sky.utils import schemas
@@ -76,6 +78,9 @@ class Kubernetes(clouds.Cloud):
                                                              'tiers are not '
                                                              'supported in '
                                                              'Kubernetes.',
+        clouds.CloudImplementationFeatures.CUSTOM_MULTI_NETWORK:
+            ('Customized multiple network interfaces are not supported in '
+             'Kubernetes.'),
         clouds.CloudImplementationFeatures.CUSTOM_NETWORK_TIER:
             ('Custom network tier is currently not supported in '
              f'{_REPR}.'),
@@ -281,7 +286,8 @@ class Kubernetes(clouds.Cloud):
                 default_value=None)
             if (autoscaler_type is not None and
                     not kubernetes_utils.get_autoscaler(
-                        autoscaler_type).can_query_backend):
+                        kubernetes_enums.KubernetesAutoscalerType(
+                            autoscaler_type)).can_query_backend):
                 # Unsupported autoscaler type. Rely on the autoscaler to
                 # provision the right instance type without running checks.
                 # Worst case, if autoscaling fails, the pod will be stuck in
@@ -298,7 +304,8 @@ class Kubernetes(clouds.Cloud):
 
             if autoscaler_type is None:
                 continue
-            autoscaler = kubernetes_utils.get_autoscaler(autoscaler_type)
+            autoscaler = kubernetes_utils.get_autoscaler(
+                kubernetes_enums.KubernetesAutoscalerType(autoscaler_type))
             logger.debug(f'{context} has autoscaler of type: {autoscaler_type}')
             if autoscaler.can_create_new_instance_of_type(
                     context, instance_type):
@@ -406,8 +413,10 @@ class Kubernetes(clouds.Cloud):
 
     @staticmethod
     def _calculate_provision_timeout(
-            num_nodes: int,
-            volume_mounts: Optional[List['volume_lib.VolumeMount']]) -> int:
+        num_nodes: int,
+        volume_mounts: Optional[List['volume_lib.VolumeMount']],
+        enable_flex_start: bool,
+    ) -> int:
         """Calculate provision timeout based on number of nodes.
 
         The timeout scales linearly with the number of nodes to account for
@@ -415,6 +424,8 @@ class Kubernetes(clouds.Cloud):
 
         Args:
             num_nodes: Number of nodes being provisioned
+            volume_mounts: Volume mounts for the pod
+            enable_flex_start: Whether flex start is enabled
 
         Returns:
             Timeout in seconds
@@ -422,7 +433,12 @@ class Kubernetes(clouds.Cloud):
         base_timeout = 10  # Base timeout for single node
         per_node_timeout = 0.2  # Additional seconds per node
         max_timeout = 60  # Cap at 1 minute
-        if volume_mounts is not None:
+        if enable_flex_start:
+            # Flex start takes longer to provision.
+            base_timeout = 1200
+            per_node_timeout = 10
+            max_timeout = 2400
+        elif volume_mounts is not None:
             for volume_mount in volume_mounts:
                 if (volume_mount.volume_config.type ==
                         volume_lib.VolumeType.PVC.value):
@@ -581,24 +597,6 @@ class Kubernetes(clouds.Cloud):
         if resources.use_spot:
             spot_label_key, spot_label_value = kubernetes_utils.get_spot_label()
 
-        # Timeout for resource provisioning. This timeout determines how long to
-        # wait for pod to be in pending status before giving up.
-        # Larger timeout may be required for autoscaling clusters, since
-        # autoscaler may take some time to provision new nodes.
-        # Note that this timeout includes time taken by the Kubernetes scheduler
-        # itself, which can be upto 2-3 seconds, and up to 10-15 seconds when
-        # scheduling 100s of pods.
-        # We use a linear scaling formula to determine the timeout based on the
-        # number of nodes.
-
-        timeout = self._calculate_provision_timeout(num_nodes, volume_mounts)
-        timeout = skypilot_config.get_effective_region_config(
-            cloud='kubernetes',
-            region=context,
-            keys=('provision_timeout',),
-            default_value=timeout,
-            override_configs=resources.cluster_config_overrides)
-
         # Check if this cluster supports high performance networking and
         # configure IPC_LOCK capability for clusters like Nebius that support it
         k8s_ipc_lock_capability = False
@@ -640,7 +638,46 @@ class Kubernetes(clouds.Cloud):
                 cloud='kubernetes',
                 region=context,
                 keys=('kueue', 'local_queue_name'),
-                default_value=None))
+                default_value=None,
+                override_configs=resources.cluster_config_overrides))
+
+        # Check DWS configuration for GKE.
+        (enable_flex_start, enable_flex_start_queued_provisioning,
+         max_run_duration_seconds) = gcp_utils.get_dws_config(
+             context, k8s_kueue_local_queue_name,
+             resources.cluster_config_overrides)
+        if enable_flex_start_queued_provisioning or enable_flex_start:
+            # DWS is only supported in GKE, check the autoscaler type.
+            autoscaler_type = skypilot_config.get_effective_region_config(
+                cloud='kubernetes',
+                region=context,
+                keys=('autoscaler',),
+                default_value=None)
+            if (autoscaler_type !=
+                    kubernetes_enums.KubernetesAutoscalerType.GKE.value):
+                raise ValueError(
+                    f'DWS is only supported in GKE, but the autoscaler type '
+                    f'for context {context} is {autoscaler_type}')
+
+        # Timeout for resource provisioning. This timeout determines how long to
+        # wait for pod to be in pending status before giving up.
+        # Larger timeout may be required for autoscaling clusters, since
+        # autoscaler may take some time to provision new nodes.
+        # Note that this timeout includes time taken by the Kubernetes scheduler
+        # itself, which can be upto 2-3 seconds, and up to 10-15 seconds when
+        # scheduling 100s of pods.
+        # We use a linear scaling formula to determine the timeout based on the
+        # number of nodes.
+
+        timeout = self._calculate_provision_timeout(
+            num_nodes, volume_mounts, enable_flex_start or
+            enable_flex_start_queued_provisioning)
+        timeout = skypilot_config.get_effective_region_config(
+            cloud='kubernetes',
+            region=context,
+            keys=('provision_timeout',),
+            default_value=timeout,
+            override_configs=resources.cluster_config_overrides)
 
         deploy_vars = {
             'instance_type': resources.instance_type,
@@ -696,6 +733,8 @@ class Kubernetes(clouds.Cloud):
                 (k8s_ha_storage_class_name),
             'avoid_label_keys': avoid_label_keys,
             'k8s_ipc_lock_capability': k8s_ipc_lock_capability,
+            'k8s_enable_flex_start': enable_flex_start,
+            'k8s_max_run_duration_seconds': max_run_duration_seconds,
         }
 
         # Add kubecontext if it is set. It may be None if SkyPilot is running

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/clouds/lambda_cloud.py

@@ -49,6 +49,9 @@ class Lambda(clouds.Cloud):
              f'{_REPR}.'),
         clouds.CloudImplementationFeatures.HOST_CONTROLLERS: f'Host controllers are not supported in {_REPR}.',
         clouds.CloudImplementationFeatures.HIGH_AVAILABILITY_CONTROLLERS: f'High availability controllers are not supported on {_REPR}.',
+        clouds.CloudImplementationFeatures.CUSTOM_MULTI_NETWORK:
+            ('Customized multiple network interfaces are not supported in '
+             f'{_REPR}.'),
     }
 
     PROVISIONER_VERSION = clouds.ProvisionerVersion.SKYPILOT

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/clouds/nebius.py

@@ -62,6 +62,9 @@ class Nebius(clouds.Cloud):
             ('Custom network tier is currently not supported on Nebius.'),
         clouds.CloudImplementationFeatures.HIGH_AVAILABILITY_CONTROLLERS:
             ('High availability controllers are not supported on Nebius.'),
+        clouds.CloudImplementationFeatures.CUSTOM_MULTI_NETWORK:
+            ('Customized multiple network interfaces are not supported on '
+             f'{_REPR}.'),
     }
     # Nebius maximum instance name length defined as <= 63 as a hostname length
     # 63 - 8 - 5 = 50 characters since

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/clouds/oci.py

@@ -80,6 +80,9 @@ class OCI(clouds.Cloud):
             clouds.CloudImplementationFeatures.HIGH_AVAILABILITY_CONTROLLERS:
                 ('High availability controllers are not supported on '
                  f'{cls._REPR}.'),
+            clouds.CloudImplementationFeatures.CUSTOM_MULTI_NETWORK:
+                ('Customized multiple network interfaces are not supported on '
+                 f'{cls._REPR}.'),
         }
         if resources.use_spot:
             unsupported_features[clouds.CloudImplementationFeatures.STOP] = (

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/clouds/paperspace.py

@@ -47,6 +47,9 @@ class Paperspace(clouds.Cloud):
              f'{_REPR}.'),
         clouds.CloudImplementationFeatures.HIGH_AVAILABILITY_CONTROLLERS:
             (f'High availability controllers are not supported in {_REPR}.'),
+        clouds.CloudImplementationFeatures.CUSTOM_MULTI_NETWORK:
+            ('Customized multiple network interfaces are not supported in '
+             f'{_REPR}.'),
     }
     _MAX_CLUSTER_NAME_LEN_LIMIT = 120
     _regions: List[clouds.Region] = []

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/clouds/runpod.py

@@ -39,6 +39,9 @@ class RunPod(clouds.Cloud):
              'to local disk.'),
         clouds.CloudImplementationFeatures.HIGH_AVAILABILITY_CONTROLLERS:
             ('High availability controllers are not supported on RunPod.'),
+        clouds.CloudImplementationFeatures.CUSTOM_MULTI_NETWORK:
+            ('Customized multiple network interfaces are not supported on '
+             'RunPod.'),
     }
     _MAX_CLUSTER_NAME_LEN_LIMIT = 120
     _regions: List[clouds.Region] = []

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/clouds/scp.py

@@ -62,6 +62,9 @@ class SCP(clouds.Cloud):
              f'{_REPR}.'),
         clouds.CloudImplementationFeatures.HIGH_AVAILABILITY_CONTROLLERS:
             (f'High availability controllers are not supported on {_REPR}.'),
+        clouds.CloudImplementationFeatures.CUSTOM_MULTI_NETWORK:
+            ('Customized multiple network interfaces are not supported on '
+             f'{_REPR}.'),
     }
 
     _INDENT_PREFIX = '    '

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/clouds/utils/gcp_utils.py

@@ -10,7 +10,7 @@ import dataclasses
 import json
 import time
 import typing
-from typing import List, Optional, Set
+from typing import Any, Dict, List, Optional, Set, Tuple
 
 import cachetools
 
@@ -18,6 +18,7 @@ from sky import sky_logging
 from sky import skypilot_config
 from sky.provision.gcp import constants
 from sky.provision.kubernetes import utils as kubernetes_utils
+from sky.utils import resources_utils
 from sky.utils import subprocess_utils
 
 if typing.TYPE_CHECKING:
@@ -206,3 +207,62 @@ def get_minimal_storage_permissions() -> List[str]:
     permissions += constants.GCP_MINIMAL_PERMISSIONS
 
     return permissions
+
+
+# Get the DWS configuration for the given context in GKE.
+def get_dws_config(
+    context: str,
+    k8s_kueue_local_queue_name: Optional[str],
+    cluster_config_overrides: Optional[Dict[str, Any]] = None,
+) -> Tuple[bool, bool, Optional[int]]:
+    """Get the DWS configuration for the given context.
+
+        Args:
+            context: The context to get the DWS configuration for.
+            k8s_kueue_local_queue_name: The name of the Kueue local queue.
+            cluster_config_overrides: The cluster config overrides.
+
+        Returns:
+            A tuple of (enable_flex_start,
+                        enable_flex_start_queued_provisioning,
+                        max_run_duration_seconds).
+
+        Raises:
+            ValueError: If k8s_kueue_local_queue_name is missing to enable
+                        flex start queued provisioning for the given context.
+        """
+    dws_config = skypilot_config.get_effective_region_config(
+        cloud='kubernetes',
+        region=context,
+        keys=('dws',),
+        default_value={},
+        override_configs=cluster_config_overrides)
+    if not dws_config:
+        return False, False, None
+
+    enabled = dws_config.get('enabled', False)
+    if not enabled:
+        return False, False, None
+
+    enable_flex_start = False
+    enable_flex_start_queued_provisioning = False
+    max_run_duration_seconds = None
+    # If users already use Kueue, use the flex start with queued
+    # provisioning mode.
+    if k8s_kueue_local_queue_name:
+        enable_flex_start_queued_provisioning = True
+    else:
+        enable_flex_start = True
+
+    if not enable_flex_start_queued_provisioning:
+        return (enable_flex_start, enable_flex_start_queued_provisioning,
+                max_run_duration_seconds)
+
+    # Max run duration is only used in the flex start with queued
+    # provisioning mode.
+    max_run_duration = dws_config.get('max_run_duration', None)
+    if max_run_duration:
+        max_run_duration_seconds = resources_utils.parse_time_minutes(
+            max_run_duration) * 60
+    return (enable_flex_start, enable_flex_start_queued_provisioning,
+            max_run_duration_seconds)

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/clouds/vast.py

@@ -35,6 +35,9 @@ class Vast(clouds.Cloud):
             ('Mounting object stores is not supported on Vast.'),
         clouds.CloudImplementationFeatures.HIGH_AVAILABILITY_CONTROLLERS:
             ('High availability controllers are not supported on Vast.'),
+        clouds.CloudImplementationFeatures.CUSTOM_MULTI_NETWORK:
+            ('Customized multiple network interfaces are not supported on Vast.'
+            ),
     }
     #
     # Vast doesn't have a max cluster name limit. This number

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/clouds/vsphere.py

@@ -60,6 +60,9 @@ class Vsphere(clouds.Cloud):
             (f'Opening ports is currently not supported on {_REPR}.'),
         clouds.CloudImplementationFeatures.HIGH_AVAILABILITY_CONTROLLERS:
             (f'High availability controllers are not supported on {_REPR}.'),
+        clouds.CloudImplementationFeatures.CUSTOM_MULTI_NETWORK:
+            (f'Customized multiple network interfaces '
+             f'are not supported on {_REPR}.'),
     }
 
     _MAX_CLUSTER_NAME_LEN_LIMIT = 80  # The name can't exceeds 80 characters

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/dashboard/out/404.html

@@ -1 +1 @@
-<!DOCTYPE html><html><head><meta charSet="utf-8"/><meta name="viewport" content="width=device-width"/><meta name="next-head-count" content="2"/><link rel="preload" href="/dashboard/_next/static/css/0da6afe66176678a.css" as="style"/><link rel="stylesheet" href="/dashboard/_next/static/css/0da6afe66176678a.css" data-n-g=""/><noscript data-n-css=""></noscript><script defer="" nomodule="" src="/dashboard/_next/static/chunks/polyfills-78c92fac7aa8fdd8.js"></script><script src="/dashboard/_next/static/chunks/webpack-9a81ea998672c303.js" defer=""></script><script src="/dashboard/_next/static/chunks/framework-efc06c2733009cd3.js" defer=""></script><script src="/dashboard/_next/static/chunks/main-c0a4f1ea606d48d2.js" defer=""></script><script src="/dashboard/_next/static/chunks/pages/_app-a37b06ddb64521fd.js" defer=""></script><script src="/dashboard/_next/static/chunks/pages/_error-c72a1f77a3c0be1b.js" defer=""></script><script src="/dashboard/_next/static/wEkAg9F21A-COXJLf20VU/_buildManifest.js" defer=""></script><script src="/dashboard/_next/static/wEkAg9F21A-COXJLf20VU/_ssgManifest.js" defer=""></script></head><body><div id="__next"></div><script id="__NEXT_DATA__" type="application/json">{"props":{"pageProps":{"statusCode":404}},"page":"/_error","query":{},"buildId":"wEkAg9F21A-COXJLf20VU","assetPrefix":"/dashboard","nextExport":true,"isFallback":false,"gip":true,"scriptLoader":[]}</script></body></html>
+<!DOCTYPE html><html><head><meta charSet="utf-8"/><meta name="viewport" content="width=device-width"/><meta name="next-head-count" content="2"/><link rel="preload" href="/dashboard/_next/static/css/0da6afe66176678a.css" as="style"/><link rel="stylesheet" href="/dashboard/_next/static/css/0da6afe66176678a.css" data-n-g=""/><noscript data-n-css=""></noscript><script defer="" nomodule="" src="/dashboard/_next/static/chunks/polyfills-78c92fac7aa8fdd8.js"></script><script src="/dashboard/_next/static/chunks/webpack-9a81ea998672c303.js" defer=""></script><script src="/dashboard/_next/static/chunks/framework-efc06c2733009cd3.js" defer=""></script><script src="/dashboard/_next/static/chunks/main-c0a4f1ea606d48d2.js" defer=""></script><script src="/dashboard/_next/static/chunks/pages/_app-a37b06ddb64521fd.js" defer=""></script><script src="/dashboard/_next/static/chunks/pages/_error-c72a1f77a3c0be1b.js" defer=""></script><script src="/dashboard/_next/static/EqELoF4IXcALfWVihInou/_buildManifest.js" defer=""></script><script src="/dashboard/_next/static/EqELoF4IXcALfWVihInou/_ssgManifest.js" defer=""></script></head><body><div id="__next"></div><script id="__NEXT_DATA__" type="application/json">{"props":{"pageProps":{"statusCode":404}},"page":"/_error","query":{},"buildId":"EqELoF4IXcALfWVihInou","assetPrefix":"/dashboard","nextExport":true,"isFallback":false,"gip":true,"scriptLoader":[]}</script></body></html>

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/dashboard/out/clusters/[cluster]/[job].html

@@ -1 +1 @@
-<!DOCTYPE html><html><head><meta charSet="utf-8"/><meta name="viewport" content="width=device-width"/><meta name="next-head-count" content="2"/><link rel="preload" href="/dashboard/_next/static/css/0da6afe66176678a.css" as="style"/><link rel="stylesheet" href="/dashboard/_next/static/css/0da6afe66176678a.css" data-n-g=""/><noscript data-n-css=""></noscript><script defer="" nomodule="" src="/dashboard/_next/static/chunks/polyfills-78c92fac7aa8fdd8.js"></script><script src="/dashboard/_next/static/chunks/webpack-9a81ea998672c303.js" defer=""></script><script src="/dashboard/_next/static/chunks/framework-efc06c2733009cd3.js" defer=""></script><script src="/dashboard/_next/static/chunks/main-c0a4f1ea606d48d2.js" defer=""></script><script src="/dashboard/_next/static/chunks/pages/_app-a37b06ddb64521fd.js" defer=""></script><script src="/dashboard/_next/static/chunks/616-162f3033ffcd3d31.js" defer=""></script><script src="/dashboard/_next/static/chunks/5230-df791914b54d91d9.js" defer=""></script><script src="/dashboard/_next/static/chunks/5739-5ea3ffa10fc884f2.js" defer=""></script><script src="/dashboard/_next/static/chunks/1664-d65361e92b85e786.js" defer=""></script><script src="/dashboard/_next/static/chunks/804-9f5e98ce84d46bdd.js" defer=""></script><script src="/dashboard/_next/static/chunks/6989-6ff4e45dfb49d11d.js" defer=""></script><script src="/dashboard/_next/static/chunks/3698-52ad1ca228faa776.js" defer=""></script><script src="/dashboard/_next/static/chunks/9470-21d059a1dfa03f61.js" defer=""></script><script src="/dashboard/_next/static/chunks/pages/clusters/%5Bcluster%5D/%5Bjob%5D-8135aba0712bda37.js" defer=""></script><script src="/dashboard/_next/static/wEkAg9F21A-COXJLf20VU/_buildManifest.js" defer=""></script><script src="/dashboard/_next/static/wEkAg9F21A-COXJLf20VU/_ssgManifest.js" defer=""></script></head><body><div id="__next"></div><script id="__NEXT_DATA__" type="application/json">{"props":{"pageProps":{}},"page":"/clusters/[cluster]/[job]","query":{},"buildId":"wEkAg9F21A-COXJLf20VU","assetPrefix":"/dashboard","nextExport":true,"autoExport":true,"isFallback":false,"scriptLoader":[]}</script></body></html>
+<!DOCTYPE html><html><head><meta charSet="utf-8"/><meta name="viewport" content="width=device-width"/><meta name="next-head-count" content="2"/><link rel="preload" href="/dashboard/_next/static/css/0da6afe66176678a.css" as="style"/><link rel="stylesheet" href="/dashboard/_next/static/css/0da6afe66176678a.css" data-n-g=""/><noscript data-n-css=""></noscript><script defer="" nomodule="" src="/dashboard/_next/static/chunks/polyfills-78c92fac7aa8fdd8.js"></script><script src="/dashboard/_next/static/chunks/webpack-9a81ea998672c303.js" defer=""></script><script src="/dashboard/_next/static/chunks/framework-efc06c2733009cd3.js" defer=""></script><script src="/dashboard/_next/static/chunks/main-c0a4f1ea606d48d2.js" defer=""></script><script src="/dashboard/_next/static/chunks/pages/_app-a37b06ddb64521fd.js" defer=""></script><script src="/dashboard/_next/static/chunks/616-162f3033ffcd3d31.js" defer=""></script><script src="/dashboard/_next/static/chunks/5230-df791914b54d91d9.js" defer=""></script><script src="/dashboard/_next/static/chunks/5739-5ea3ffa10fc884f2.js" defer=""></script><script src="/dashboard/_next/static/chunks/1664-d65361e92b85e786.js" defer=""></script><script src="/dashboard/_next/static/chunks/804-9f5e98ce84d46bdd.js" defer=""></script><script src="/dashboard/_next/static/chunks/6989-6ff4e45dfb49d11d.js" defer=""></script><script src="/dashboard/_next/static/chunks/3698-52ad1ca228faa776.js" defer=""></script><script src="/dashboard/_next/static/chunks/9470-21d059a1dfa03f61.js" defer=""></script><script src="/dashboard/_next/static/chunks/pages/clusters/%5Bcluster%5D/%5Bjob%5D-8135aba0712bda37.js" defer=""></script><script src="/dashboard/_next/static/EqELoF4IXcALfWVihInou/_buildManifest.js" defer=""></script><script src="/dashboard/_next/static/EqELoF4IXcALfWVihInou/_ssgManifest.js" defer=""></script></head><body><div id="__next"></div><script id="__NEXT_DATA__" type="application/json">{"props":{"pageProps":{}},"page":"/clusters/[cluster]/[job]","query":{},"buildId":"EqELoF4IXcALfWVihInou","assetPrefix":"/dashboard","nextExport":true,"autoExport":true,"isFallback":false,"scriptLoader":[]}</script></body></html>

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/dashboard/out/clusters/[cluster].html

@@ -1 +1 @@
-<!DOCTYPE html><html><head><meta charSet="utf-8"/><meta name="viewport" content="width=device-width"/><meta name="next-head-count" content="2"/><link rel="preload" href="/dashboard/_next/static/css/0da6afe66176678a.css" as="style"/><link rel="stylesheet" href="/dashboard/_next/static/css/0da6afe66176678a.css" data-n-g=""/><noscript data-n-css=""></noscript><script defer="" nomodule="" src="/dashboard/_next/static/chunks/polyfills-78c92fac7aa8fdd8.js"></script><script src="/dashboard/_next/static/chunks/webpack-9a81ea998672c303.js" defer=""></script><script src="/dashboard/_next/static/chunks/framework-efc06c2733009cd3.js" defer=""></script><script src="/dashboard/_next/static/chunks/main-c0a4f1ea606d48d2.js" defer=""></script><script src="/dashboard/_next/static/chunks/pages/_app-a37b06ddb64521fd.js" defer=""></script><script src="/dashboard/_next/static/chunks/616-162f3033ffcd3d31.js" defer=""></script><script src="/dashboard/_next/static/chunks/5230-df791914b54d91d9.js" defer=""></script><script src="/dashboard/_next/static/chunks/5739-5ea3ffa10fc884f2.js" defer=""></script><script src="/dashboard/_next/static/chunks/1664-d65361e92b85e786.js" defer=""></script><script src="/dashboard/_next/static/chunks/804-9f5e98ce84d46bdd.js" defer=""></script><script src="/dashboard/_next/static/chunks/1272-1ef0bf0237faccdb.js" defer=""></script><script src="/dashboard/_next/static/chunks/3947-b059261d6fa88a1f.js" defer=""></script><script src="/dashboard/_next/static/chunks/6989-6ff4e45dfb49d11d.js" defer=""></script><script src="/dashboard/_next/static/chunks/3698-52ad1ca228faa776.js" defer=""></script><script src="/dashboard/_next/static/chunks/9470-21d059a1dfa03f61.js" defer=""></script><script src="/dashboard/_next/static/chunks/6990-d0dc765474fa0eca.js" defer=""></script><script src="/dashboard/_next/static/chunks/8969-909d53833da080cb.js" defer=""></script><script src="/dashboard/_next/static/chunks/1043-1b39779691bb4030.js" defer=""></script><script src="/dashboard/_next/static/chunks/6601-fcfad0ddf92ec7ab.js" defer=""></script><script src="/dashboard/_next/static/chunks/938-044ad21de8b4626b.js" defer=""></script><script src="/dashboard/_next/static/chunks/1871-80dea41717729fa5.js" defer=""></script><script src="/dashboard/_next/static/chunks/pages/clusters/%5Bcluster%5D-1159f362b960e2b8.js" defer=""></script><script src="/dashboard/_next/static/wEkAg9F21A-COXJLf20VU/_buildManifest.js" defer=""></script><script src="/dashboard/_next/static/wEkAg9F21A-COXJLf20VU/_ssgManifest.js" defer=""></script></head><body><div id="__next"></div><script id="__NEXT_DATA__" type="application/json">{"props":{"pageProps":{}},"page":"/clusters/[cluster]","query":{},"buildId":"wEkAg9F21A-COXJLf20VU","assetPrefix":"/dashboard","nextExport":true,"autoExport":true,"isFallback":false,"scriptLoader":[]}</script></body></html>
+<!DOCTYPE html><html><head><meta charSet="utf-8"/><meta name="viewport" content="width=device-width"/><meta name="next-head-count" content="2"/><link rel="preload" href="/dashboard/_next/static/css/0da6afe66176678a.css" as="style"/><link rel="stylesheet" href="/dashboard/_next/static/css/0da6afe66176678a.css" data-n-g=""/><noscript data-n-css=""></noscript><script defer="" nomodule="" src="/dashboard/_next/static/chunks/polyfills-78c92fac7aa8fdd8.js"></script><script src="/dashboard/_next/static/chunks/webpack-9a81ea998672c303.js" defer=""></script><script src="/dashboard/_next/static/chunks/framework-efc06c2733009cd3.js" defer=""></script><script src="/dashboard/_next/static/chunks/main-c0a4f1ea606d48d2.js" defer=""></script><script src="/dashboard/_next/static/chunks/pages/_app-a37b06ddb64521fd.js" defer=""></script><script src="/dashboard/_next/static/chunks/616-162f3033ffcd3d31.js" defer=""></script><script src="/dashboard/_next/static/chunks/5230-df791914b54d91d9.js" defer=""></script><script src="/dashboard/_next/static/chunks/5739-5ea3ffa10fc884f2.js" defer=""></script><script src="/dashboard/_next/static/chunks/1664-d65361e92b85e786.js" defer=""></script><script src="/dashboard/_next/static/chunks/804-9f5e98ce84d46bdd.js" defer=""></script><script src="/dashboard/_next/static/chunks/1272-1ef0bf0237faccdb.js" defer=""></script><script src="/dashboard/_next/static/chunks/3947-b059261d6fa88a1f.js" defer=""></script><script src="/dashboard/_next/static/chunks/6989-6ff4e45dfb49d11d.js" defer=""></script><script src="/dashboard/_next/static/chunks/3698-52ad1ca228faa776.js" defer=""></script><script src="/dashboard/_next/static/chunks/9470-21d059a1dfa03f61.js" defer=""></script><script src="/dashboard/_next/static/chunks/6990-d0dc765474fa0eca.js" defer=""></script><script src="/dashboard/_next/static/chunks/8969-909d53833da080cb.js" defer=""></script><script src="/dashboard/_next/static/chunks/1043-1b39779691bb4030.js" defer=""></script><script src="/dashboard/_next/static/chunks/6601-fcfad0ddf92ec7ab.js" defer=""></script><script src="/dashboard/_next/static/chunks/938-044ad21de8b4626b.js" defer=""></script><script src="/dashboard/_next/static/chunks/1871-80dea41717729fa5.js" defer=""></script><script src="/dashboard/_next/static/chunks/pages/clusters/%5Bcluster%5D-1159f362b960e2b8.js" defer=""></script><script src="/dashboard/_next/static/EqELoF4IXcALfWVihInou/_buildManifest.js" defer=""></script><script src="/dashboard/_next/static/EqELoF4IXcALfWVihInou/_ssgManifest.js" defer=""></script></head><body><div id="__next"></div><script id="__NEXT_DATA__" type="application/json">{"props":{"pageProps":{}},"page":"/clusters/[cluster]","query":{},"buildId":"EqELoF4IXcALfWVihInou","assetPrefix":"/dashboard","nextExport":true,"autoExport":true,"isFallback":false,"scriptLoader":[]}</script></body></html>

{skypilot_nightly-1.0.0.dev20250707 → skypilot_nightly-1.0.0.dev20250709}/sky/dashboard/out/clusters.html

@@ -1 +1 @@
-<!DOCTYPE html><html><head><meta charSet="utf-8"/><meta name="viewport" content="width=device-width"/><meta name="next-head-count" content="2"/><link rel="preload" href="/dashboard/_next/static/css/0da6afe66176678a.css" as="style"/><link rel="stylesheet" href="/dashboard/_next/static/css/0da6afe66176678a.css" data-n-g=""/><noscript data-n-css=""></noscript><script defer="" nomodule="" src="/dashboard/_next/static/chunks/polyfills-78c92fac7aa8fdd8.js"></script><script src="/dashboard/_next/static/chunks/webpack-9a81ea998672c303.js" defer=""></script><script src="/dashboard/_next/static/chunks/framework-efc06c2733009cd3.js" defer=""></script><script src="/dashboard/_next/static/chunks/main-c0a4f1ea606d48d2.js" defer=""></script><script src="/dashboard/_next/static/chunks/pages/_app-a37b06ddb64521fd.js" defer=""></script><script src="/dashboard/_next/static/chunks/pages/clusters-9744c271a1642f76.js" defer=""></script><script src="/dashboard/_next/static/wEkAg9F21A-COXJLf20VU/_buildManifest.js" defer=""></script><script src="/dashboard/_next/static/wEkAg9F21A-COXJLf20VU/_ssgManifest.js" defer=""></script></head><body><div id="__next"></div><script id="__NEXT_DATA__" type="application/json">{"props":{"pageProps":{}},"page":"/clusters","query":{},"buildId":"wEkAg9F21A-COXJLf20VU","assetPrefix":"/dashboard","nextExport":true,"autoExport":true,"isFallback":false,"scriptLoader":[]}</script></body></html>
+<!DOCTYPE html><html><head><meta charSet="utf-8"/><meta name="viewport" content="width=device-width"/><meta name="next-head-count" content="2"/><link rel="preload" href="/dashboard/_next/static/css/0da6afe66176678a.css" as="style"/><link rel="stylesheet" href="/dashboard/_next/static/css/0da6afe66176678a.css" data-n-g=""/><noscript data-n-css=""></noscript><script defer="" nomodule="" src="/dashboard/_next/static/chunks/polyfills-78c92fac7aa8fdd8.js"></script><script src="/dashboard/_next/static/chunks/webpack-9a81ea998672c303.js" defer=""></script><script src="/dashboard/_next/static/chunks/framework-efc06c2733009cd3.js" defer=""></script><script src="/dashboard/_next/static/chunks/main-c0a4f1ea606d48d2.js" defer=""></script><script src="/dashboard/_next/static/chunks/pages/_app-a37b06ddb64521fd.js" defer=""></script><script src="/dashboard/_next/static/chunks/pages/clusters-9744c271a1642f76.js" defer=""></script><script src="/dashboard/_next/static/EqELoF4IXcALfWVihInou/_buildManifest.js" defer=""></script><script src="/dashboard/_next/static/EqELoF4IXcALfWVihInou/_ssgManifest.js" defer=""></script></head><body><div id="__next"></div><script id="__NEXT_DATA__" type="application/json">{"props":{"pageProps":{}},"page":"/clusters","query":{},"buildId":"EqELoF4IXcALfWVihInou","assetPrefix":"/dashboard","nextExport":true,"autoExport":true,"isFallback":false,"scriptLoader":[]}</script></body></html>