skypilot-nightly 1.0.0.dev20241014__py3-none-any.whl → 1.0.0.dev20241016__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 = '340f38404fe5d3ebe35ea430a67cb3377241d1f3'
+_SKYPILOT_COMMIT_SHA = '53380e26f01452559012d57b333b17f40dd8a4d1'
 
 
 def _get_git_commit():
@@ -35,7 +35,7 @@ def _get_git_commit():
 
 
 __commit__ = _get_git_commit()
-__version__ = '1.0.0.dev20241014'
+__version__ = '1.0.0.dev20241016'
 __root_dir__ = os.path.dirname(os.path.abspath(__file__))
 
 

sky/backends/backend_utils.py

@@ -56,7 +56,7 @@ from sky.utils import timeline
 from sky.utils import ux_utils
 
 if typing.TYPE_CHECKING:
-    from sky import resources
+    from sky import resources as resources_lib
     from sky import task as task_lib
     from sky.backends import cloud_vm_ray_backend
     from sky.backends import local_docker_backend
@@ -751,7 +751,7 @@ def _replace_yaml_dicts(
 # TODO: too many things happening here - leaky abstraction. Refactor.
 @timeline.event
 def write_cluster_config(
-        to_provision: 'resources.Resources',
+        to_provision: 'resources_lib.Resources',
         num_nodes: int,
         cluster_config_template: str,
         cluster_name: str,

sky/backends/cloud_vm_ray_backend.py

@@ -2844,9 +2844,9 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
                         time.sleep(gap_seconds)
                         continue
                     logger.error(
-                        f'{colorama.Fore.RED}⨯{colorama.Style.RESET_ALL} '
-                        'Failed to provision resources. '
-                        f'{ux_utils.log_path_hint(log_path)}')
+                        ux_utils.error_message(
+                            'Failed to provision resources. '
+                            f'{ux_utils.log_path_hint(log_path)}'))
                     error_message += (
                         '\nTo keep retrying until the cluster is up, use '
                         'the `--retry-until-up` flag.')

sky/cli.py

@@ -1464,54 +1464,8 @@ def _status_kubernetes(show_all: bool):
     Args:
         show_all (bool): Show all job information (e.g., start time, failures).
     """
-    context = kubernetes_utils.get_current_kube_config_context_name()
-    try:
-        pods = kubernetes_utils.get_skypilot_pods(context)
-    except exceptions.ResourcesUnavailableError as e:
-        with ux_utils.print_exception_no_traceback():
-            raise ValueError('Failed to get SkyPilot pods from '
-                             f'Kubernetes: {str(e)}') from e
-    all_clusters, jobs_controllers, serve_controllers = (
-        status_utils.process_skypilot_pods(pods, context))
-    all_jobs = []
-    with rich_utils.safe_status(
-            '[bold cyan]Checking in-progress managed jobs[/]') as spinner:
-        for i, (_, job_controller_info) in enumerate(jobs_controllers.items()):
-            user = job_controller_info['user']
-            pod = job_controller_info['pods'][0]
-            status_message = ('[bold cyan]Checking managed jobs controller')
-            if len(jobs_controllers) > 1:
-                status_message += f's ({i+1}/{len(jobs_controllers)})'
-            spinner.update(f'{status_message}[/]')
-            try:
-                job_list = managed_jobs.queue_from_kubernetes_pod(
-                    pod.metadata.name)
-            except RuntimeError as e:
-                logger.warning('Failed to get managed jobs from controller '
-                               f'{pod.metadata.name}: {str(e)}')
-                job_list = []
-            # Add user field to jobs
-            for job in job_list:
-                job['user'] = user
-            all_jobs.extend(job_list)
-    # Reconcile cluster state between managed jobs and clusters:
-    # To maintain a clear separation between regular SkyPilot clusters
-    # and those from managed jobs, we need to exclude the latter from
-    # the main cluster list.
-    # We do this by reconstructing managed job cluster names from each
-    # job's name and ID. We then use this set to filter out managed
-    # clusters from the main cluster list. This is necessary because there
-    # are no identifiers distinguishing clusters from managed jobs from
-    # regular clusters.
-    managed_job_cluster_names = set()
-    for job in all_jobs:
-        # Managed job cluster name is <job_name>-<job_id>
-        managed_cluster_name = f'{job["job_name"]}-{job["job_id"]}'
-        managed_job_cluster_names.add(managed_cluster_name)
-    unmanaged_clusters = [
-        c for c in all_clusters
-        if c['cluster_name'] not in managed_job_cluster_names
-    ]
+    all_clusters, unmanaged_clusters, all_jobs, context = (
+        core.status_kubernetes())
     click.echo(f'{colorama.Fore.CYAN}{colorama.Style.BRIGHT}'
                f'Kubernetes cluster state (context: {context})'
                f'{colorama.Style.RESET_ALL}')
@@ -1523,7 +1477,7 @@ def _status_kubernetes(show_all: bool):
                    f'{colorama.Style.RESET_ALL}')
         msg = managed_jobs.format_job_table(all_jobs, show_all=show_all)
         click.echo(msg)
-    if serve_controllers:
+    if any(['sky-serve-controller' in c.cluster_name for c in all_clusters]):
         # TODO: Parse serve controllers and show services separately.
         #  Currently we show a hint that services are shown as clusters.
         click.echo(f'\n{colorama.Style.DIM}Hint: SkyServe replica pods are '

sky/clouds/gcp.py

@@ -94,6 +94,12 @@ _IMAGE_NOT_FOUND_UX_MESSAGE = (
     f'\nTo query common AI images: {colorama.Style.BRIGHT}gcloud compute images list --project deeplearning-platform-release | less{colorama.Style.RESET_ALL}'
 )
 
+# Image ID tags
+_DEFAULT_CPU_IMAGE_ID = 'skypilot:custom-cpu-ubuntu-2204'
+# For GPU-related package version, see sky/clouds/service_catalog/images/provisioners/cuda.sh
+_DEFAULT_GPU_IMAGE_ID = 'skypilot:custom-gpu-ubuntu-2204'
+_DEFAULT_GPU_K80_IMAGE_ID = 'skypilot:k80-debian-10'
+
 
 def _run_output(cmd):
     proc = subprocess.run(cmd,
@@ -422,7 +428,7 @@ class GCP(clouds.Cloud):
         # --no-standard-images
         # We use the debian image, as the ubuntu image has some connectivity
         # issue when first booted.
-        image_id = 'skypilot:cpu-debian-11'
+        image_id = _DEFAULT_CPU_IMAGE_ID
 
         def _failover_disk_tier() -> Optional[resources_utils.DiskTier]:
             if (r.disk_tier is not None and
@@ -487,10 +493,10 @@ class GCP(clouds.Cloud):
                     # Though the image is called cu113, it actually has later
                     # versions of CUDA as noted below.
                     # CUDA driver version 470.57.02, CUDA Library 11.4
-                    image_id = 'skypilot:k80-debian-10'
+                    image_id = _DEFAULT_GPU_K80_IMAGE_ID
                 else:
                     # CUDA driver version 535.86.10, CUDA Library 12.2
-                    image_id = 'skypilot:gpu-debian-11'
+                    image_id = _DEFAULT_GPU_IMAGE_ID
 
         if (resources.image_id is not None and
                 resources.extract_docker_image() is None):

sky/core.py

@@ -1,7 +1,7 @@
 """SDK functions for cluster/job management."""
 import getpass
 import typing
-from typing import Any, Dict, List, Optional, Union
+from typing import Any, Dict, List, Optional, Tuple, Union
 
 import colorama
 
@@ -11,10 +11,12 @@ from sky import dag
 from sky import data
 from sky import exceptions
 from sky import global_user_state
+from sky import jobs as managed_jobs
 from sky import sky_logging
 from sky import status_lib
 from sky import task
 from sky.backends import backend_utils
+from sky.provision.kubernetes import utils as kubernetes_utils
 from sky.skylet import constants
 from sky.skylet import job_lib
 from sky.usage import usage_lib
@@ -111,6 +113,79 @@ def status(cluster_names: Optional[Union[str, List[str]]] = None,
                                       cluster_names=cluster_names)
 
 
+def status_kubernetes(
+) -> Tuple[List['kubernetes_utils.KubernetesSkyPilotClusterInfo'],
+           List['kubernetes_utils.KubernetesSkyPilotClusterInfo'], List[Dict[
+               str, Any]], Optional[str]]:
+    """Get all SkyPilot clusters and jobs in the Kubernetes cluster.
+
+    Managed jobs and services are also included in the clusters returned.
+    The caller must parse the controllers to identify which clusters are run
+    as managed jobs or services.
+all_clusters, unmanaged_clusters, all_jobs, context
+    Returns:
+        A tuple containing:
+        - all_clusters: List of KubernetesSkyPilotClusterInfo with info for
+            all clusters, including managed jobs, services and controllers.
+        - unmanaged_clusters: List of KubernetesSkyPilotClusterInfo with info
+            for all clusters excluding managed jobs and services. Controllers
+            are included.
+        - all_jobs: List of managed jobs from all controllers. Each entry is a
+            dictionary job info, see jobs.queue_from_kubernetes_pod for details.
+        - context: Kubernetes context used to fetch the cluster information.
+    """
+    context = kubernetes_utils.get_current_kube_config_context_name()
+    try:
+        pods = kubernetes_utils.get_skypilot_pods(context)
+    except exceptions.ResourcesUnavailableError as e:
+        with ux_utils.print_exception_no_traceback():
+            raise ValueError('Failed to get SkyPilot pods from '
+                             f'Kubernetes: {str(e)}') from e
+    all_clusters, jobs_controllers, _ = (kubernetes_utils.process_skypilot_pods(
+        pods, context))
+    all_jobs = []
+    with rich_utils.safe_status(
+            ux_utils.spinner_message(
+                '[bold cyan]Checking in-progress managed jobs[/]')) as spinner:
+        for i, job_controller_info in enumerate(jobs_controllers):
+            user = job_controller_info.user
+            pod = job_controller_info.pods[0]
+            status_message = '[bold cyan]Checking managed jobs controller'
+            if len(jobs_controllers) > 1:
+                status_message += f's ({i + 1}/{len(jobs_controllers)})'
+            spinner.update(f'{status_message}[/]')
+            try:
+                job_list = managed_jobs.queue_from_kubernetes_pod(
+                    pod.metadata.name)
+            except RuntimeError as e:
+                logger.warning('Failed to get managed jobs from controller '
+                               f'{pod.metadata.name}: {str(e)}')
+                job_list = []
+            # Add user field to jobs
+            for job in job_list:
+                job['user'] = user
+            all_jobs.extend(job_list)
+    # Reconcile cluster state between managed jobs and clusters:
+    # To maintain a clear separation between regular SkyPilot clusters
+    # and those from managed jobs, we need to exclude the latter from
+    # the main cluster list.
+    # We do this by reconstructing managed job cluster names from each
+    # job's name and ID. We then use this set to filter out managed
+    # clusters from the main cluster list. This is necessary because there
+    # are no identifiers distinguishing clusters from managed jobs from
+    # regular clusters.
+    managed_job_cluster_names = set()
+    for job in all_jobs:
+        # Managed job cluster name is <job_name>-<job_id>
+        managed_cluster_name = f'{job["job_name"]}-{job["job_id"]}'
+        managed_job_cluster_names.add(managed_cluster_name)
+    unmanaged_clusters = [
+        c for c in all_clusters
+        if c.cluster_name not in managed_job_cluster_names
+    ]
+    return all_clusters, unmanaged_clusters, all_jobs, context
+
+
 def endpoints(cluster: str,
               port: Optional[Union[int, str]] = None) -> Dict[int, str]:
     """Gets the endpoint for a given cluster and port number (endpoint).

sky/provision/kubernetes/utils.py

@@ -15,9 +15,11 @@ import jinja2
 import yaml
 
 import sky
+from sky import clouds
 from sky import exceptions
 from sky import sky_logging
 from sky import skypilot_config
+from sky import status_lib
 from sky.adaptors import kubernetes
 from sky.provision import constants as provision_constants
 from sky.provision.kubernetes import network_utils
@@ -30,6 +32,7 @@ from sky.utils import ux_utils
 
 if typing.TYPE_CHECKING:
     from sky import backends
+    from sky import resources as resources_lib
 
 # TODO(romilb): Move constants to constants.py
 DEFAULT_NAMESPACE = 'default'
@@ -2023,3 +2026,113 @@ def get_skypilot_pods(context: Optional[str] = None) -> List[Any]:
             'kubectl get pods --selector=skypilot-cluster --all-namespaces'
         ) from None
     return pods
+
+
+@dataclasses.dataclass
+class KubernetesSkyPilotClusterInfo:
+    cluster_name_on_cloud: str
+    cluster_name: str
+    user: str
+    status: status_lib.ClusterStatus
+    pods: List[Any]
+    launched_at: float
+    resources: 'resources_lib.Resources'
+    resources_str: str
+
+
+def process_skypilot_pods(
+    pods: List[Any],
+    context: Optional[str] = None
+) -> Tuple[List[KubernetesSkyPilotClusterInfo],
+           List[KubernetesSkyPilotClusterInfo],
+           List[KubernetesSkyPilotClusterInfo]]:
+    """Process SkyPilot pods on k8s to extract cluster and controller info.
+
+    Args:
+        pods: List of Kubernetes pod objects.
+        context: Kubernetes context name, used to detect GPU label formatter.
+
+    Returns:
+        A tuple containing:
+        - List of KubernetesSkyPilotClusterInfo with all cluster info.
+        - List of KubernetesSkyPilotClusterInfo with job controller info.
+        - List of KubernetesSkyPilotClusterInfo with serve controller info.
+    """
+    # pylint: disable=import-outside-toplevel
+    from sky import resources as resources_lib
+    clusters: Dict[str, KubernetesSkyPilotClusterInfo] = {}
+    jobs_controllers: List[KubernetesSkyPilotClusterInfo] = []
+    serve_controllers: List[KubernetesSkyPilotClusterInfo] = []
+
+    for pod in pods:
+        cluster_name_on_cloud = pod.metadata.labels.get('skypilot-cluster')
+        cluster_name = cluster_name_on_cloud.rsplit(
+            '-', 1
+        )[0]  # Remove the user hash to get cluster name (e.g., mycluster-2ea4)
+        if cluster_name_on_cloud not in clusters:
+            # Parse the start time for the cluster
+            start_time = pod.status.start_time
+            if start_time is not None:
+                start_time = pod.status.start_time.timestamp()
+
+            # Parse resources
+            cpu_request = parse_cpu_or_gpu_resource(
+                pod.spec.containers[0].resources.requests.get('cpu', '0'))
+            memory_request = parse_memory_resource(
+                pod.spec.containers[0].resources.requests.get('memory', '0'),
+                unit='G')
+            gpu_count = parse_cpu_or_gpu_resource(
+                pod.spec.containers[0].resources.requests.get(
+                    'nvidia.com/gpu', '0'))
+            gpu_name = None
+            if gpu_count > 0:
+                label_formatter, _ = (detect_gpu_label_formatter(context))
+                assert label_formatter is not None, (
+                    'GPU label formatter cannot be None if there are pods '
+                    f'requesting GPUs: {pod.metadata.name}')
+                gpu_label = label_formatter.get_label_key()
+                # Get GPU name from pod node selector
+                if pod.spec.node_selector is not None:
+                    gpu_name = label_formatter.get_accelerator_from_label_value(
+                        pod.spec.node_selector.get(gpu_label))
+
+            resources = resources_lib.Resources(
+                cloud=clouds.Kubernetes(),
+                cpus=int(cpu_request),
+                memory=int(memory_request),
+                accelerators=(f'{gpu_name}:{gpu_count}'
+                              if gpu_count > 0 else None))
+            if pod.status.phase == 'Pending':
+                # If pod is pending, do not show it in the status
+                continue
+
+            cluster_info = KubernetesSkyPilotClusterInfo(
+                cluster_name_on_cloud=cluster_name_on_cloud,
+                cluster_name=cluster_name,
+                user=pod.metadata.labels.get('skypilot-user'),
+                status=status_lib.ClusterStatus.UP,
+                pods=[],
+                launched_at=start_time,
+                resources=resources,
+                resources_str='')
+            clusters[cluster_name_on_cloud] = cluster_info
+            # Check if cluster name is name of a controller
+            # Can't use controller_utils.Controllers.from_name(cluster_name)
+            # because hash is different across users
+            if 'sky-jobs-controller' in cluster_name_on_cloud:
+                jobs_controllers.append(cluster_info)
+            elif 'sky-serve-controller' in cluster_name_on_cloud:
+                serve_controllers.append(cluster_info)
+        else:
+            # Update start_time if this pod started earlier
+            pod_start_time = pod.status.start_time
+            if pod_start_time is not None:
+                pod_start_time = pod_start_time.timestamp()
+                if pod_start_time < clusters[cluster_name_on_cloud].launched_at:
+                    clusters[cluster_name_on_cloud].launched_at = pod_start_time
+        clusters[cluster_name_on_cloud].pods.append(pod)
+    # Update resources_str in clusters:
+    for cluster in clusters.values():
+        num_pods = len(cluster.pods)
+        cluster.resources_str = f'{num_pods}x {cluster.resources}'
+    return list(clusters.values()), jobs_controllers, serve_controllers

sky/provision/provisioner.py

@@ -571,7 +571,10 @@ def post_provision_runtime_setup(
                                          provision_record=provision_record,
                                          custom_resource=custom_resource)
         except Exception:  # pylint: disable=broad-except
-            logger.error('*** Failed setting up cluster. ***')
+            logger.error(
+                ux_utils.error_message(
+                    'Failed to set up SkyPilot runtime on cluster.',
+                    provision_logging.config.log_path))
             logger.debug(f'Stacktrace:\n{traceback.format_exc()}')
             with ux_utils.print_exception_no_traceback():
                 raise

sky/skylet/constants.py

@@ -155,8 +155,8 @@ CONDA_INSTALLATION_COMMANDS = (
     # We use --system-site-packages to reuse the system site packages to avoid
     # the overhead of installing the same packages in the new environment.
     f'[ -d {SKY_REMOTE_PYTHON_ENV} ] || '
-    f'{{ {SKY_PYTHON_CMD} -m venv {SKY_REMOTE_PYTHON_ENV} --system-site-packages && '
-    f'echo "$(echo {SKY_REMOTE_PYTHON_ENV})/bin/python" > {SKY_PYTHON_PATH_FILE}; }};'
+    f'{SKY_PYTHON_CMD} -m venv {SKY_REMOTE_PYTHON_ENV} --system-site-packages;'
+    f'echo "$(echo {SKY_REMOTE_PYTHON_ENV})/bin/python" > {SKY_PYTHON_PATH_FILE};'
 )
 
 _sky_version = str(version.parse(sky.__version__))

sky/utils/cli_utils/status_utils.py

@@ -1,19 +1,20 @@
 """Utilities for sky status."""
-from typing import Any, Callable, Dict, List, Optional, Tuple
+import typing
+from typing import Any, Callable, Dict, List, Optional
 
 import click
 import colorama
 
 from sky import backends
-from sky import clouds as sky_clouds
-from sky import resources as resources_lib
 from sky import status_lib
-from sky.provision.kubernetes import utils as kubernetes_utils
 from sky.skylet import constants
 from sky.utils import common_utils
 from sky.utils import log_utils
 from sky.utils import resources_utils
 
+if typing.TYPE_CHECKING:
+    from sky.provision.kubernetes import utils as kubernetes_utils
+
 COMMAND_TRUNC_LENGTH = 25
 NUM_COST_REPORT_LINES = 5
 
@@ -303,19 +304,19 @@ def _get_estimated_cost_for_cost_report(
     return f'$ {cost:.2f}'
 
 
-def show_kubernetes_cluster_status_table(clusters: List[Any],
-                                         show_all: bool) -> None:
+def show_kubernetes_cluster_status_table(
+        clusters: List['kubernetes_utils.KubernetesSkyPilotClusterInfo'],
+        show_all: bool) -> None:
     """Compute cluster table values and display for Kubernetes clusters."""
     status_columns = [
-        StatusColumn('USER', lambda c: c['user']),
-        StatusColumn('NAME', lambda c: c['cluster_name']),
-        StatusColumn(
-            'LAUNCHED',
-            lambda c: log_utils.readable_time_duration(c['launched_at'])),
+        StatusColumn('USER', lambda c: c.user),
+        StatusColumn('NAME', lambda c: c.cluster_name),
+        StatusColumn('LAUNCHED',
+                     lambda c: log_utils.readable_time_duration(c.launched_at)),
         StatusColumn('RESOURCES',
-                     lambda c: c['resources_str'],
+                     lambda c: c.resources_str,
                      trunc_length=70 if not show_all else 0),
-        StatusColumn('STATUS', lambda c: c['status'].colored_str()),
+        StatusColumn('STATUS', lambda c: c.status.colored_str()),
         # TODO(romilb): We should consider adding POD_NAME field here when --all
         #  is passed to help users fetch pod name programmatically.
     ]
@@ -326,8 +327,7 @@ def show_kubernetes_cluster_status_table(clusters: List[Any],
     cluster_table = log_utils.create_table(columns)
 
     # Sort table by user, then by cluster name
-    sorted_clusters = sorted(clusters,
-                             key=lambda c: (c['user'], c['cluster_name']))
+    sorted_clusters = sorted(clusters, key=lambda c: (c.user, c.cluster_name))
 
     for cluster in sorted_clusters:
         row = []
@@ -344,122 +344,3 @@ def show_kubernetes_cluster_status_table(clusters: List[Any],
     else:
         click.echo('No SkyPilot resources found in the '
                    'active Kubernetes context.')
-
-
-def process_skypilot_pods(
-    pods: List[Any],
-    context: Optional[str] = None
-) -> Tuple[List[Dict[Any, Any]], Dict[str, Any], Dict[str, Any]]:
-    """Process SkyPilot pods on k8s to extract cluster and controller info.
-
-    Args:
-        pods: List of Kubernetes pod objects.
-        context: Kubernetes context name, used to detect GPU label formatter.
-
-    Returns:
-        A tuple containing:
-        - List of dictionaries with cluster information.
-        - Dictionary of job controller information.
-        - Dictionary of serve controller information.
-
-        Each dictionary contains the following keys:
-            'cluster_name_on_cloud': The cluster_name_on_cloud used by SkyPilot
-            'cluster_name': The cluster name without the user hash
-            'user': The user who created the cluster. Fetched from pod label
-            'status': The cluster status (assumed UP if pod exists)
-            'pods': List of pod objects in the cluster
-            'launched_at': Timestamp of when the cluster was launched
-            'resources': sky.Resources object for the cluster
-    """
-    clusters: Dict[str, Dict] = {}
-    jobs_controllers: Dict[str, Dict] = {}
-    serve_controllers: Dict[str, Dict] = {}
-
-    for pod in pods:
-        cluster_name_on_cloud = pod.metadata.labels.get('skypilot-cluster')
-        cluster_name = cluster_name_on_cloud.rsplit(
-            '-', 1
-        )[0]  # Remove the user hash to get cluster name (e.g., mycluster-2ea4)
-
-        # Check if cluster name is name of a controller
-        # Can't use controller_utils.Controllers.from_name(cluster_name)
-        # because hash is different across users
-        if 'controller' in cluster_name_on_cloud:
-            start_time = pod.status.start_time.timestamp()
-            controller_info = {
-                'cluster_name_on_cloud': cluster_name_on_cloud,
-                'cluster_name': cluster_name,
-                'user': pod.metadata.labels.get('skypilot-user'),
-                'status': status_lib.ClusterStatus.UP,
-                # Assuming UP if pod exists
-                'pods': [pod],
-                'launched_at': start_time
-            }
-            if 'sky-jobs-controller' in cluster_name_on_cloud:
-                jobs_controllers[cluster_name_on_cloud] = controller_info
-            elif 'sky-serve-controller' in cluster_name_on_cloud:
-                serve_controllers[cluster_name_on_cloud] = controller_info
-
-        if cluster_name_on_cloud not in clusters:
-            # Parse the start time for the cluster
-            start_time = pod.status.start_time
-            if start_time is not None:
-                start_time = pod.status.start_time.timestamp()
-
-            # Parse resources
-            cpu_request = kubernetes_utils.parse_cpu_or_gpu_resource(
-                pod.spec.containers[0].resources.requests.get('cpu', '0'))
-            memory_request = kubernetes_utils.parse_memory_resource(
-                pod.spec.containers[0].resources.requests.get('memory', '0'),
-                unit='G')
-            gpu_count = kubernetes_utils.parse_cpu_or_gpu_resource(
-                pod.spec.containers[0].resources.requests.get(
-                    'nvidia.com/gpu', '0'))
-            if gpu_count > 0:
-                label_formatter, _ = (
-                    kubernetes_utils.detect_gpu_label_formatter(context))
-                assert label_formatter is not None, (
-                    'GPU label formatter cannot be None if there are pods '
-                    f'requesting GPUs: {pod.metadata.name}')
-                gpu_label = label_formatter.get_label_key()
-                # Get GPU name from pod node selector
-                if pod.spec.node_selector is not None:
-                    gpu_name = label_formatter.get_accelerator_from_label_value(
-                        pod.spec.node_selector.get(gpu_label))
-
-            resources = resources_lib.Resources(
-                cloud=sky_clouds.Kubernetes(),
-                cpus=int(cpu_request),
-                memory=int(memory_request),
-                accelerators=(f'{gpu_name}:{gpu_count}'
-                              if gpu_count > 0 else None))
-            if pod.status.phase == 'Pending':
-                # If pod is pending, do not show it in the status
-                continue
-
-            clusters[cluster_name_on_cloud] = {
-                'cluster_name_on_cloud': cluster_name_on_cloud,
-                'cluster_name': cluster_name,
-                'user': pod.metadata.labels.get('skypilot-user'),
-                'status': status_lib.ClusterStatus.UP,
-                'pods': [],
-                'launched_at': start_time,
-                'resources': resources,
-            }
-        else:
-            # Update start_time if this pod started earlier
-            pod_start_time = pod.status.start_time
-            if pod_start_time is not None:
-                pod_start_time = pod_start_time.timestamp()
-                if pod_start_time < clusters[cluster_name_on_cloud][
-                        'launched_at']:
-                    clusters[cluster_name_on_cloud][
-                        'launched_at'] = pod_start_time
-        clusters[cluster_name_on_cloud]['pods'].append(pod)
-    # Update resources_str in clusters:
-    for cluster_name, cluster in clusters.items():
-        resources = cluster['resources']
-        num_pods = len(cluster['pods'])
-        resources_str = f'{num_pods}x {resources}'
-        cluster['resources_str'] = resources_str
-    return list(clusters.values()), jobs_controllers, serve_controllers

sky/utils/ux_utils.py

@@ -121,11 +121,6 @@ class RedirectOutputForProcess:
                 raise
 
 
-def starting_message(message: str) -> str:
-    """Gets the starting message for the given message."""
-    return f'⚙︎ {message}'
-
-
 def log_path_hint(log_path: Union[str, 'pathlib.Path']) -> str:
     """Gets the log path hint for the given log path."""
     log_path = str(log_path)
@@ -135,21 +130,50 @@ def log_path_hint(log_path: Union[str, 'pathlib.Path']) -> str:
     return _LOG_PATH_HINT.format(log_path=log_path)
 
 
+def starting_message(message: str) -> str:
+    """Gets the starting message for the given message."""
+    # We have to reset the color before the message, because sometimes if a
+    # previous spinner with dimmed color overflows in a narrow terminal, the
+    # color might be messed up.
+    return f'{colorama.Style.RESET_ALL}⚙︎ {message}'
+
+
 def finishing_message(
         message: str,
         log_path: Optional[Union[str, 'pathlib.Path']] = None) -> str:
     """Gets the finishing message for the given message."""
-    success_prefix = (f'{colorama.Fore.GREEN}✓ {message}'
-                      f'{colorama.Style.RESET_ALL}')
+    # We have to reset the color before the message, because sometimes if a
+    # previous spinner with dimmed color overflows in a narrow terminal, the
+    # color might be messed up.
+    success_prefix = (f'{colorama.Style.RESET_ALL}{colorama.Fore.GREEN}✓ '
+                      f'{message}{colorama.Style.RESET_ALL}')
     if log_path is None:
         return success_prefix
     path_hint = log_path_hint(log_path)
     return f'{success_prefix}  {path_hint}'
 
 
+def error_message(message: str,
+                  log_path: Optional[Union[str, 'pathlib.Path']] = None) -> str:
+    """Gets the error message for the given message."""
+    # We have to reset the color before the message, because sometimes if a
+    # previous spinner with dimmed color overflows in a narrow terminal, the
+    # color might be messed up.
+    error_prefix = (f'{colorama.Style.RESET_ALL}{colorama.Fore.RED}⨯'
+                    f'{colorama.Style.RESET_ALL} {message}')
+    if log_path is None:
+        return error_prefix
+    path_hint = log_path_hint(log_path)
+    return f'{error_prefix}  {path_hint}'
+
+
 def retry_message(message: str) -> str:
     """Gets the retry message for the given message."""
-    return f'{colorama.Fore.YELLOW}↺{colorama.Style.RESET_ALL} {message}'
+    # We have to reset the color before the message, because sometimes if a
+    # previous spinner with dimmed color overflows in a narrow terminal, the
+    # color might be messed up.
+    return (f'{colorama.Style.RESET_ALL}{colorama.Fore.YELLOW}↺'
+            f'{colorama.Style.RESET_ALL} {message}')
 
 
 def spinner_message(

{skypilot_nightly-1.0.0.dev20241014.dist-info → skypilot_nightly-1.0.0.dev20241016.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: skypilot-nightly
-Version: 1.0.0.dev20241014
+Version: 1.0.0.dev20241016
 Summary: SkyPilot: An intercloud broker for the clouds
 Author: SkyPilot Team
 License: Apache 2.0

{skypilot_nightly-1.0.0.dev20241014.dist-info → skypilot_nightly-1.0.0.dev20241016.dist-info}/RECORD

@@ -1,10 +1,10 @@
-sky/__init__.py,sha256=VNwdB81dfCeF4uRD4BuPNbQ3MmE9-P1PX9hLxEzmWmw,5854
+sky/__init__.py,sha256=19EG_Nr4EJcbkyLvfF_ZmWhAbfEysS498RvephjOslM,5854
 sky/admin_policy.py,sha256=hPo02f_A32gCqhUueF0QYy1fMSSKqRwYEg_9FxScN_s,3248
 sky/authentication.py,sha256=TfKkVnmRIetATSEVQFp-rOOIRGqVig2i8faSQQt_ixA,20974
 sky/check.py,sha256=jLMIIJrseaZj1_o5WkbaD9XdyXIlCaT6pyAaIFdhdmA,9079
-sky/cli.py,sha256=eC-uC2_OhuDephCdejr70VeB1bo1KGwHC8et5VR_lto,210828
+sky/cli.py,sha256=PJR6W92twf89j17OWLQJ9RawdazJcGslfW2L_fLB2PM,208545
 sky/cloud_stores.py,sha256=RjFgmRhUh1Kk__f6g3KxzLp9s7dA0pFK4W1AukEuUaw,21153
-sky/core.py,sha256=I7ZxjrJmrM-zAWVnmUyz0kzS9fjKaaLehDdWMm3H0x8,34443
+sky/core.py,sha256=DW9OGE2kS2CmsvQ1grrpRnNFS3woMGWSHu5GE99e-I4,38190
 sky/dag.py,sha256=WLFWr5hfrwjd31uYlNvI-zWUk7tLaT_gzJn4LzbVtkE,2780
 sky/exceptions.py,sha256=D7WARzYRt4dGjXo6gI-gzkoodZbKF1D-qncm_DbHB28,8846
 sky/execution.py,sha256=CbrKMgfc2JgLqZqwPvmYKxbWAQKYqHpOLpUEOb-k2m0,24679
@@ -30,8 +30,8 @@ sky/adaptors/runpod.py,sha256=4Nt_BfZhJAKQNA3wO8cxvvNI8x4NsDGHu_4EhRDlGYQ,225
 sky/adaptors/vsphere.py,sha256=zJP9SeObEoLrpgHW2VHvZE48EhgVf8GfAEIwBeaDMfM,2129
 sky/backends/__init__.py,sha256=UDjwbUgpTRApbPJnNfR786GadUuwgRk3vsWoVu5RB_c,536
 sky/backends/backend.py,sha256=wwfbrxPhjMPs6PSyy3tAHI8WJhl-xhgzWBsAZjmJJ6g,6249
-sky/backends/backend_utils.py,sha256=Kp3HRqjrisnDYzLURYSbtyZ7WGMKHSVfMSwRj5_aOnc,126770
-sky/backends/cloud_vm_ray_backend.py,sha256=otDjYA4FtnlFGU_xNaxhBM5wFQ8fuF4kPeYQmf1Me3E,236901
+sky/backends/backend_utils.py,sha256=u9P7Fd3DB9LaOq51fK7kwKpxtgFGGWmgULY6GoLSUPM,126791
+sky/backends/cloud_vm_ray_backend.py,sha256=9mCLLRUD-x3ksiiPbhrMDsZWIPNU9cVSQwwpmxSia7k,236881
 sky/backends/docker_utils.py,sha256=Hyw1YY20EyghhEbYx6O2FIMDcGkNzBzV9TM7LFynei8,8358
 sky/backends/local_docker_backend.py,sha256=0JL5m0YUgOmOL4aWEUe4tmt89dsxjk4_WXkPwgEKEis,16801
 sky/backends/wheel_utils.py,sha256=3QS4T_Ydvo4DbYhogtyADyNBEf04I6jUCL71M285shQ,7963
@@ -46,7 +46,7 @@ sky/clouds/cloud.py,sha256=PPk-Cbf1YbJT8bswcQLtPBtko02OWrRGJKkLzDpytTI,34858
 sky/clouds/cloud_registry.py,sha256=4yQMv-iBSgyN5aNL4Qxbn0JVE-dkVoEUIgj7S1z9S_Q,955
 sky/clouds/cudo.py,sha256=H4VyMo5wWGAv2MXZ3xsbWjlZA_cZYnt4ecNlTOOao8Y,13147
 sky/clouds/fluidstack.py,sha256=iOmoOx52yTrHKMzwBDaxFJCfNo79M61d5tj-Np24Lyc,12436
-sky/clouds/gcp.py,sha256=CrSsaSXd83tM78foKH9viBfW1cQsjve3aUQbshsqvDg,54033
+sky/clouds/gcp.py,sha256=FKHqtF4YMY06pseloMEbnt4XwIQ5ErDLlrvyXzIzZa4,54308
 sky/clouds/ibm.py,sha256=M8QdjeSFlwssfoY2aOodxG4q5R3eT9K-4lTPDHYvEYI,21476
 sky/clouds/kubernetes.py,sha256=aWoXWR-S4puZHzuUHroLKxLdTpkqU7j75dQlXECnsmE,28679
 sky/clouds/lambda_cloud.py,sha256=2Al3qCSl-I4iTi7pPPNXcbaLyVfCUgTl__vYBunLB6k,12439
@@ -110,7 +110,7 @@ sky/provision/docker_utils.py,sha256=Z7vDUs9Yjqks_CsWrACcTgABIZuFi3EJVFwkU0WsdD0
 sky/provision/instance_setup.py,sha256=n1Px_KOYZl7Rf1WLXrfTTHyqxyA8_5QTN9BNLjQRkgc,22427
 sky/provision/logging.py,sha256=yZWgejrFBhhRjAtvFu5N5bRXIMK5TuwNjp1vKQqz2pw,2103
 sky/provision/metadata_utils.py,sha256=LrxeV4wD2QPzNdXV_npj8q-pr35FatxBBjF_jSbpOT0,4013
-sky/provision/provisioner.py,sha256=7BatzBobHGCLJf2u5hCt4VveS9tOPc5zL4wGHR6OcqE,24888
+sky/provision/provisioner.py,sha256=A4-yY0Q4GnkdJsHl_DLNEycq5wFKFsPwT0fwTNh1dG0,25016
 sky/provision/aws/__init__.py,sha256=mxq8PeWJqUtalDozTNpbtENErRZ1ktEs8uf2aG9UUgU,731
 sky/provision/aws/config.py,sha256=ApEh63RR_KyCp9nPXX35z6jBREoulJPQ5st8K9Jlclo,23385
 sky/provision/aws/instance.py,sha256=eCslJ2XfJo_pkQMnKFQqhGnUIRvwKiT12oxBY5-klss,40750
@@ -141,7 +141,7 @@ sky/provision/kubernetes/config.py,sha256=WEKcFXXhe89bLGAvoMiBvTDxdxkpTIA6ezrj2v
 sky/provision/kubernetes/instance.py,sha256=Qth9AWc8OBGB7WeGJ4ERlopNA8y2wg1AvS5XPJEuXXQ,38421
 sky/provision/kubernetes/network.py,sha256=EpNjRQ131CXepqbdkoRKFu4szVrm0oKEpv1l8EgOkjU,12364
 sky/provision/kubernetes/network_utils.py,sha256=t1FS3K400fetH7cBuRgQJZl5_jEeMshsvsYmnMUcq8k,11399
-sky/provision/kubernetes/utils.py,sha256=oJgCrbR8IyTw_uMoM9oxYYYNN_tZ1yzppWNTDM-XqaM,84522
+sky/provision/kubernetes/utils.py,sha256=2N5c4yA7CEn4DjvCiUO73W4XDEjgixcJRVdgs913QQE,89523
 sky/provision/kubernetes/manifests/smarter-device-manager-configmap.yaml,sha256=AMzYzlY0JIlfBWj5eX054Rc1XDW2thUcLSOGMJVhIdA,229
 sky/provision/kubernetes/manifests/smarter-device-manager-daemonset.yaml,sha256=RtTq4F1QUmR2Uunb6zuuRaPhV7hpesz4saHjn3Ncsb4,2010
 sky/provision/paperspace/__init__.py,sha256=1nbUPWio7UA5gCQkO_rfEDfgXT17u5OtuByxQx4Ez6g,598
@@ -187,7 +187,7 @@ sky/skylet/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 sky/skylet/attempt_skylet.py,sha256=GZ6ITjjA0m-da3IxXXfoHR6n4pjp3X3TOXUqVvSrV0k,2136
 sky/skylet/autostop_lib.py,sha256=JPDHmByuhoNYXSUHl-OnyeJUkOFWn7gDM1FrS7Kr3E8,4478
 sky/skylet/configs.py,sha256=UtnpmEL0F9hH6PSjhsps7xgjGZ6qzPOfW1p2yj9tSng,1887
-sky/skylet/constants.py,sha256=txAK0602kGtaD42JUYECq5u4rLIZFhOIWz-fLUV1KgA,14652
+sky/skylet/constants.py,sha256=OsuJcQp6UgkQ9Yfml6f_raXXbHS7-_h-v4QNv92y0Gw,14642
 sky/skylet/events.py,sha256=A09E7LmmwzcGrSG0n8K7d3EZ1ZJr1mmmzoGyhnArYJA,12303
 sky/skylet/job_lib.py,sha256=csXhJ6lvAxjzSy2PzZvr0sPJyEoVkCvPi2Zjq9JZkHY,35884
 sky/skylet/log_lib.py,sha256=7rdmEr5Wy9CIGLkWls5-FVeqTL1O2GPy44uDaEAqMS0,19312
@@ -258,10 +258,10 @@ sky/utils/rich_utils.py,sha256=hmnI1X5dKvRIQzB7EyNb34FT97qFNve-0QHqM5r0mVk,3066
 sky/utils/schemas.py,sha256=QT0Fxri2o0SiWkky1DlZhA1dzQRQoB5OdVaej0wJvhc,28787
 sky/utils/subprocess_utils.py,sha256=3R54Elc2n8DQeO6Y8MCDJ6N6v27HDGpbNMIfCquqXYQ,6552
 sky/utils/timeline.py,sha256=ao_nm0y52ZQILfL7Y92c3pSEFRyPm_ElORC3DrI5BwQ,3936
-sky/utils/ux_utils.py,sha256=0Fy20UbM2rQ_aVIx_B2-oiLXGmzC1myn-0pnmb56pcA,5227
+sky/utils/ux_utils.py,sha256=CqyIFGDuSE8fQasPkna_loZMwtboC9KedR09WEQ7qz0,6502
 sky/utils/validator.py,sha256=cAFERCoC7jH0DFKepcU4x9SYmdrYL1iVmW9tXA18hvo,701
 sky/utils/cli_utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-sky/utils/cli_utils/status_utils.py,sha256=zIaiGWtEACdczubiSBMTupueZStPt1VqgRrhQilLEaI,17954
+sky/utils/cli_utils/status_utils.py,sha256=2HrH6IBJCJ__AbuZ0ooIEgarBKIVIA5M3myE5qYvToU,12330
 sky/utils/kubernetes/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 sky/utils/kubernetes/create_cluster.sh,sha256=VLXfazav9XCMQmeKVqhuOQzt2vM6G1jgnvvb0SHUFno,7773
 sky/utils/kubernetes/delete_cluster.sh,sha256=BSccHF43GyepDNf-FZcenzHzpXXATkVD92vgn1lWPgk,927
@@ -273,9 +273,9 @@ sky/utils/kubernetes/k8s_gpu_labeler_job.yaml,sha256=KPqp23B-zQ2SZK03jdHeF9fLTog
 sky/utils/kubernetes/k8s_gpu_labeler_setup.yaml,sha256=VLKT2KKimZu1GDg_4AIlIt488oMQvhRZWwsj9vBbPUg,3812
 sky/utils/kubernetes/rsync_helper.sh,sha256=aRMa_0JRHtXFOPtEg4rFAwR1t57wvvAoGZhn3H3BtGk,1059
 sky/utils/kubernetes/ssh_jump_lifecycle_manager.py,sha256=RFLJ3k7MR5UN4SKHykQ0lV9SgXumoULpKYIAt1vh-HU,6560
-skypilot_nightly-1.0.0.dev20241014.dist-info/LICENSE,sha256=emRJAvE7ngL6x0RhQvlns5wJzGI3NEQ_WMjNmd9TZc4,12170
-skypilot_nightly-1.0.0.dev20241014.dist-info/METADATA,sha256=oTglaVnp9O-PgRAk74UFb1rMIKKjm9z896VnhSIV83g,18945
-skypilot_nightly-1.0.0.dev20241014.dist-info/WHEEL,sha256=GV9aMThwP_4oNCtvEC2ec3qUYutgWeAzklro_0m4WJQ,91
-skypilot_nightly-1.0.0.dev20241014.dist-info/entry_points.txt,sha256=StA6HYpuHj-Y61L2Ze-hK2IcLWgLZcML5gJu8cs6nU4,36
-skypilot_nightly-1.0.0.dev20241014.dist-info/top_level.txt,sha256=qA8QuiNNb6Y1OF-pCUtPEr6sLEwy2xJX06Bd_CrtrHY,4
-skypilot_nightly-1.0.0.dev20241014.dist-info/RECORD,,
+skypilot_nightly-1.0.0.dev20241016.dist-info/LICENSE,sha256=emRJAvE7ngL6x0RhQvlns5wJzGI3NEQ_WMjNmd9TZc4,12170
+skypilot_nightly-1.0.0.dev20241016.dist-info/METADATA,sha256=3cBwGMlr5S-mHHm8ZXtnMUNjCcSYeAzwOEf-N4LxLEU,18945
+skypilot_nightly-1.0.0.dev20241016.dist-info/WHEEL,sha256=OVMc5UfuAQiSplgO0_WdW7vXVGAt9Hdd6qtN4HotdyA,91
+skypilot_nightly-1.0.0.dev20241016.dist-info/entry_points.txt,sha256=StA6HYpuHj-Y61L2Ze-hK2IcLWgLZcML5gJu8cs6nU4,36
+skypilot_nightly-1.0.0.dev20241016.dist-info/top_level.txt,sha256=qA8QuiNNb6Y1OF-pCUtPEr6sLEwy2xJX06Bd_CrtrHY,4
+skypilot_nightly-1.0.0.dev20241016.dist-info/RECORD,,

{skypilot_nightly-1.0.0.dev20241014.dist-info → skypilot_nightly-1.0.0.dev20241016.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (75.1.0)
+Generator: setuptools (75.2.0)
 Root-Is-Purelib: true
 Tag: py3-none-any