skypilot-nightly 1.0.0.dev20241110__py3-none-any.whl → 1.0.0.dev20241112__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 = 'dddd65187953a5d6b32f762bea78eed1f109ec3c'
+_SKYPILOT_COMMIT_SHA = '140125eaad5fb64da37934c8f6650d68aa135f77'
 
 
 def _get_git_commit():
@@ -35,7 +35,7 @@ def _get_git_commit():
 
 
 __commit__ = _get_git_commit()
-__version__ = '1.0.0.dev20241110'
+__version__ = '1.0.0.dev20241112'
 __root_dir__ = os.path.dirname(os.path.abspath(__file__))
 
 

sky/cli.py

@@ -3102,6 +3102,7 @@ def show_gpus(
     kubernetes_autoscaling = kubernetes_utils.get_autoscaler_type() is not None
     kubernetes_is_enabled = sky_clouds.cloud_in_iterable(
         sky_clouds.Kubernetes(), global_user_state.get_cached_enabled_clouds())
+    no_permissions_str = '<no permissions>'
 
     def _list_to_str(lst):
         return ', '.join([str(e) for e in lst])
@@ -3146,9 +3147,11 @@ def show_gpus(
                             debug_msg)
             raise ValueError(full_err_msg)
         for gpu, _ in sorted(counts.items()):
+            available_qty = available[gpu] if available[gpu] != -1 else (
+                no_permissions_str)
             realtime_gpu_table.add_row([
                 gpu,
-                _list_to_str(counts.pop(gpu)), capacity[gpu], available[gpu]
+                _list_to_str(counts.pop(gpu)), capacity[gpu], available_qty
             ])
         return realtime_gpu_table
 
@@ -3158,10 +3161,11 @@ def show_gpus(
 
         node_info_dict = kubernetes_utils.get_kubernetes_node_info(context)
         for node_name, node_info in node_info_dict.items():
+            available = node_info.free['nvidia.com/gpu'] if node_info.free[
+                'nvidia.com/gpu'] != -1 else no_permissions_str
             node_table.add_row([
                 node_name, node_info.gpu_type,
-                node_info.total['nvidia.com/gpu'],
-                node_info.free['nvidia.com/gpu']
+                node_info.total['nvidia.com/gpu'], available
             ])
         return node_table
 

sky/clouds/service_catalog/kubernetes_catalog.py

@@ -10,6 +10,7 @@ from typing import Dict, List, Optional, Set, Tuple
 from sky import check as sky_check
 from sky import sky_logging
 from sky.adaptors import common as adaptors_common
+from sky.adaptors import kubernetes
 from sky.clouds import Kubernetes
 from sky.clouds.service_catalog import CloudFilter
 from sky.clouds.service_catalog import common
@@ -22,6 +23,8 @@ if typing.TYPE_CHECKING:
 else:
     pd = adaptors_common.LazyImport('pandas')
 
+logger = sky_logging.init_logger(__name__)
+
 _PULL_FREQUENCY_HOURS = 7
 
 # We keep pull_frequency_hours so we can remotely update the default image paths
@@ -77,6 +80,11 @@ def list_accelerators_realtime(
     require_price: bool = True
 ) -> Tuple[Dict[str, List[common.InstanceTypeInfo]], Dict[str, int], Dict[str,
                                                                           int]]:
+    """List accelerators in the Kubernetes cluster.
+
+    If the user does not have sufficient permissions to list pods in all
+    namespaces, the function will return free GPUs as -1.
+    """
     # TODO(romilb): This should be refactored to use get_kubernetes_node_info()
     #   function from kubernetes_utils.
     del all_regions, require_price  # Unused.
@@ -108,7 +116,17 @@ def list_accelerators_realtime(
     key = label_formatter.get_label_key()
     nodes = kubernetes_utils.get_kubernetes_nodes(context)
     # Get the pods to get the real-time GPU usage
-    pods = kubernetes_utils.get_all_pods_in_kubernetes_cluster(context)
+    try:
+        pods = kubernetes_utils.get_all_pods_in_kubernetes_cluster(context)
+    except kubernetes.api_exception() as e:
+        if e.status == 403:
+            logger.warning('Failed to get pods in the Kubernetes cluster '
+                           '(forbidden). Please check if your account has '
+                           'necessary permissions to list pods. Realtime GPU '
+                           'availability information may be incorrect.')
+            pods = None
+        else:
+            raise
     # Total number of GPUs in the cluster
     total_accelerators_capacity: Dict[str, int] = {}
     # Total number of GPUs currently available in the cluster
@@ -141,6 +159,21 @@ def list_accelerators_realtime(
                 if accelerator_count not in accelerators_qtys:
                     accelerators_qtys.add((accelerator_name, accelerator_count))
 
+            if accelerator_count >= min_quantity_filter:
+                quantized_count = (min_quantity_filter *
+                                   (accelerator_count // min_quantity_filter))
+                if accelerator_name not in total_accelerators_capacity:
+                    total_accelerators_capacity[
+                        accelerator_name] = quantized_count
+                else:
+                    total_accelerators_capacity[
+                        accelerator_name] += quantized_count
+
+            if pods is None:
+                # If we can't get the pods, we can't get the GPU usage
+                total_accelerators_available[accelerator_name] = -1
+                continue
+
             for pod in pods:
                 # Get all the pods running on the node
                 if (pod.spec.node_name == node.metadata.name and
@@ -155,16 +188,6 @@ def list_accelerators_realtime(
 
             accelerators_available = accelerator_count - allocated_qty
 
-            if accelerator_count >= min_quantity_filter:
-                quantized_count = (min_quantity_filter *
-                                   (accelerator_count // min_quantity_filter))
-                if accelerator_name not in total_accelerators_capacity:
-                    total_accelerators_capacity[
-                        accelerator_name] = quantized_count
-                else:
-                    total_accelerators_capacity[
-                        accelerator_name] += quantized_count
-
             if accelerator_name not in total_accelerators_available:
                 total_accelerators_available[accelerator_name] = 0
             if accelerators_available >= min_quantity_filter:

sky/provision/docker_utils.py

@@ -20,7 +20,7 @@ SETUP_ENV_VARS_CMD = (
     '{ if [ $(id -u) -ne 0 ]; then echo "sudo"; else echo ""; fi; } && '
     'printenv | while IFS=\'=\' read -r key value; do echo "export $key=\\\"$value\\\""; done > '  # pylint: disable=line-too-long
     '~/container_env_var.sh && '
-    '$(prefix_cmd) mv ~/container_env_var.sh /etc/profile.d/container_env_var.sh'
+    '$(prefix_cmd) mv ~/container_env_var.sh /etc/profile.d/container_env_var.sh;'
 )
 
 # Docker daemon may not be ready when the machine is firstly started. The error

sky/provision/kubernetes/instance.py

@@ -333,52 +333,37 @@ def _run_function_with_retries(func: Callable,
                 raise
 
 
-def _set_env_vars_in_pods(namespace: str, context: Optional[str],
-                          new_pods: List):
-    """Setting environment variables in pods.
-
-    Once all containers are ready, we can exec into them and set env vars.
-    Kubernetes automatically populates containers with critical
-    environment variables, such as those for discovering services running
-    in the cluster and CUDA/nvidia environment variables. We need to
-    make sure these env vars are available in every task and ssh session.
-    This is needed for GPU support and service discovery.
-    See https://github.com/skypilot-org/skypilot/issues/2287 for
-    more details.
-
-    To do so, we capture env vars from the pod's runtime and write them to
-    /etc/profile.d/, making them available for all users in future
-    shell sessions.
-    """
-    set_k8s_env_var_cmd = docker_utils.SETUP_ENV_VARS_CMD
+def pre_init(namespace: str, context: Optional[str], new_nodes: List) -> None:
+    """Pre-initialization step for SkyPilot pods.
 
-    def _set_env_vars_thread(new_pod):
-        pod_name = new_pod.metadata.name
-        logger.info(f'{"-"*20}Start: Set up env vars in pod {pod_name!r} '
-                    f'{"-"*20}')
-        runner = command_runner.KubernetesCommandRunner(
-            ((namespace, context), pod_name))
+    This step is run in the pod right after it is created and before the
+    SkyPilot runtime is setup.
 
-        def _run_env_vars_cmd():
-            rc, stdout, _ = runner.run(set_k8s_env_var_cmd,
-                                       require_outputs=True,
-                                       stream_logs=False)
-            _raise_command_running_error('set env vars', set_k8s_env_var_cmd,
-                                         pod_name, rc, stdout)
+    This step includes three key steps:
 
-        _run_function_with_retries(_run_env_vars_cmd,
-                                   f'set env vars in pod {pod_name}')
-        logger.info(f'{"-"*20}End: Set up env vars in pod {pod_name!r} '
-                    f'{"-"*20}')
+    1. Privilege check: Checks if the default user has sufficient privilege
+    to set up the kubernetes instance pod.
+    2. SSH setup: Sets up SSH for the pod instance.
+    3. Environment variable setup to populate k8s env vars in the pod.
 
-    subprocess_utils.run_in_parallel(_set_env_vars_thread, new_pods,
-                                     NUM_THREADS)
+    Make sure commands used in these methods are generic and work
+    on most base images. E.g., do not use Python, since that may not
+    be installed by default.
 
+    If you run any apt commands, be sure to check if the lock is available.
+    It is possible the `apt update` run in the pod container args may still
+    be running.
+
+    Args:
+        namespace (str): Kubernetes namespace.
+        context (Optional[str]): Kubernetes context.
+        new_nodes (List): List of new pod instances.
+
+    Raises:
+        config_lib.KubernetesError: If user privileges are insufficient or
+          setup fails.
+    """
 
-def _check_user_privilege(namespace: str, context: Optional[str],
-                          new_nodes: List) -> None:
-    # Checks if the default user has sufficient privilege to set up
-    # the kubernetes instance pod.
     check_k8s_user_sudo_cmd = (
         'if [ $(id -u) -eq 0 ]; then'
         # If user is root, create an alias for sudo used in skypilot setup
@@ -386,56 +371,67 @@ def _check_user_privilege(namespace: str, context: Optional[str],
         'else '
         '  if command -v sudo >/dev/null 2>&1; then '
         '    timeout 2 sudo -l >/dev/null 2>&1 && echo succeed || '
-        f'    ( echo {exceptions.INSUFFICIENT_PRIVILEGES_CODE!r}; ); '
+        f'    ( echo {exceptions.INSUFFICIENT_PRIVILEGES_CODE!r}; '
+        f'      exit {exceptions.INSUFFICIENT_PRIVILEGES_CODE}; ); '
         '  else '
-        f'    ( echo {exceptions.INSUFFICIENT_PRIVILEGES_CODE!r}; ); '
+        f'    ( echo {exceptions.INSUFFICIENT_PRIVILEGES_CODE!r}; '
+        f'      exit {exceptions.INSUFFICIENT_PRIVILEGES_CODE}; ); '
         '  fi; '
-        'fi')
+        'fi;')
+
+    # Kubernetes automatically populates containers with critical
+    # environment variables, such as those for discovering services running
+    # in the cluster and CUDA/nvidia environment variables. We need to
+    # make sure these env vars are available in every task and ssh session.
+    # This is needed for GPU support and service discovery.
+    # See https://github.com/skypilot-org/skypilot/issues/2287 for more details.
+    # To do so, we capture env vars from the pod's runtime and write them to
+    # /etc/profile.d/, making them available for all users in future
+    # shell sessions.
+    set_k8s_env_var_cmd = docker_utils.SETUP_ENV_VARS_CMD
 
-    # This check needs to run on a per-image basis, so running the check on
-    # any one pod is sufficient.
-    new_node = new_nodes[0]
-    pod_name = new_node.metadata.name
+    check_apt_update_complete_cmd = (
+        'echo "Checking if apt update from container init is complete..."; '
+        'timeout_secs=600; '
+        'start_time=$(date +%s); '
+        'while ! grep -q "Fetched" /tmp/apt-update.log 2>/dev/null; do '
+        '  echo "apt update still running. Logs:"; '
+        '  cat /tmp/apt-update.log; '
+        '  current_time=$(date +%s); '
+        '  elapsed=$((current_time - start_time)); '
+        '  if [ $elapsed -ge $timeout_secs ]; then '
+        '    echo "Timed out waiting for apt update"; '
+        '    exit 1; '
+        '  fi; '
+        '  sleep 5; '
+        'done; '
+        'echo "apt update complete."; ')
 
-    runner = command_runner.KubernetesCommandRunner(
-        ((namespace, context), pod_name))
-    logger.info(f'{"-"*20}Start: Check user privilege in pod {pod_name!r} '
-                f'{"-"*20}')
-
-    def _run_privilege_check():
-        rc, stdout, stderr = runner.run(check_k8s_user_sudo_cmd,
-                                        require_outputs=True,
-                                        separate_stderr=True,
-                                        stream_logs=False)
-        _raise_command_running_error('check user privilege',
-                                     check_k8s_user_sudo_cmd, pod_name, rc,
-                                     stdout + stderr)
-        return stdout
-
-    stdout = _run_function_with_retries(
-        _run_privilege_check, f'check user privilege in pod {pod_name!r}')
-
-    if stdout == str(exceptions.INSUFFICIENT_PRIVILEGES_CODE):
-        raise config_lib.KubernetesError(
-            'Insufficient system privileges detected. '
-            'Ensure the default user has root access or '
-            '"sudo" is installed and the user is added to the sudoers '
-            'from the image.')
-    logger.info(f'{"-"*20}End: Check user privilege in pod {pod_name!r} '
-                f'{"-"*20}')
-
-
-def _setup_ssh_in_pods(namespace: str, context: Optional[str],
-                       new_nodes: List) -> None:
-    # Setting up ssh for the pod instance. This is already setup for
-    # the jump pod so it does not need to be run for it.
-    set_k8s_ssh_cmd = (
-        'set -ex; '
+    install_ssh_k8s_cmd = (
         'prefix_cmd() '
         '{ if [ $(id -u) -ne 0 ]; then echo "sudo"; else echo ""; fi; }; '
         'export DEBIAN_FRONTEND=noninteractive;'
-        '$(prefix_cmd) apt-get update;'
-        '$(prefix_cmd) apt install openssh-server rsync -y; '
+        'echo "Installing missing packages..."; '
+        'for i in {1..5}; do '
+        '  output=$($(prefix_cmd) apt install openssh-server rsync -y 2>&1); '
+        '  rc=$?; '
+        '  if [ $rc -eq 0 ]; then '
+        '    break; '
+        '  fi; '
+        '  echo "$output" | grep -qi "could not get lock" || '
+        '  grep -qi "Unable to acquire the dpkg frontend lock"; '
+        '  if [ $? -eq 0 ]; then '
+        '    echo "apt install failed due to lock, retrying. (Attempt $i/5)"; '
+        '    sleep 5; '
+        '  else '
+        '    echo "apt install failed for a non-lock reason: $output"; '
+        '    exit $rc; '
+        '  fi; '
+        'done; '
+        'if [ $rc -ne 0 ]; then '
+        '    echo "apt install failed after 5 attempts due to lock errors."; '
+        '    exit $rc; '
+        'fi; '
         '$(prefix_cmd) mkdir -p /var/run/sshd; '
         '$(prefix_cmd) '
         'sed -i "s/PermitRootLogin prohibit-password/PermitRootLogin yes/" '
@@ -456,24 +452,35 @@ def _setup_ssh_in_pods(namespace: str, context: Optional[str],
         # See https://www.educative.io/answers/error-mesg-ttyname-failed-inappropriate-ioctl-for-device  # pylint: disable=line-too-long
         '$(prefix_cmd) sed -i "s/mesg n/tty -s \\&\\& mesg n/" ~/.profile;')
 
-    def _setup_ssh_thread(new_node):
+    pre_init_cmd = ('set -ex; ' + check_k8s_user_sudo_cmd +
+                    set_k8s_env_var_cmd + check_apt_update_complete_cmd +
+                    install_ssh_k8s_cmd)
+
+    def _pre_init_thread(new_node):
         pod_name = new_node.metadata.name
+        logger.info(f'{"-"*20}Start: Pre-init in pod {pod_name!r} {"-"*20}')
         runner = command_runner.KubernetesCommandRunner(
             ((namespace, context), pod_name))
-        logger.info(f'{"-"*20}Start: Set up SSH in pod {pod_name!r} {"-"*20}')
 
-        def _run_ssh_setup():
-            rc, stdout, _ = runner.run(set_k8s_ssh_cmd,
-                                       require_outputs=True,
-                                       stream_logs=False)
-            _raise_command_running_error('setup ssh', set_k8s_ssh_cmd, pod_name,
-                                         rc, stdout)
+        # Run the combined pre-init command
+        rc, stdout, _ = runner.run(pre_init_cmd,
+                                   require_outputs=True,
+                                   stream_logs=False)
+        if rc == exceptions.INSUFFICIENT_PRIVILEGES_CODE:
+            raise config_lib.KubernetesError(
+                'Insufficient system privileges detected. '
+                'Ensure the default user has root access or '
+                '"sudo" is installed and the user is added to the sudoers '
+                'from the image.')
+
+        op_name = 'pre-init'
+        _raise_command_running_error(op_name, pre_init_cmd, pod_name, rc,
+                                     stdout)
 
-        _run_function_with_retries(_run_ssh_setup,
-                                   f'setup ssh in pod {pod_name!r}')
-        logger.info(f'{"-"*20}End: Set up SSH in pod {pod_name!r} {"-"*20}')
+        logger.info(f'{"-"*20}End: Pre-init in pod {pod_name!r} {"-"*20}')
 
-    subprocess_utils.run_in_parallel(_setup_ssh_thread, new_nodes, NUM_THREADS)
+    # Run pre_init in parallel across all new_nodes
+    subprocess_utils.run_in_parallel(_pre_init_thread, new_nodes, NUM_THREADS)
 
 
 def _label_pod(namespace: str, context: Optional[str], pod_name: str,
@@ -724,13 +731,8 @@ def _create_pods(region: str, cluster_name_on_cloud: str,
                      f'pods: {list(uninitialized_pods.keys())}')
         uninitialized_pods_list = list(uninitialized_pods.values())
 
-        # Setup SSH and environment variables in pods.
-        # Make sure commands used in these methods are generic and work
-        # on most base images. E.g., do not use Python, since that may not
-        # be installed by default.
-        _check_user_privilege(namespace, context, uninitialized_pods_list)
-        _setup_ssh_in_pods(namespace, context, uninitialized_pods_list)
-        _set_env_vars_in_pods(namespace, context, uninitialized_pods_list)
+        # Run pre-init steps in the pod.
+        pre_init(namespace, context, uninitialized_pods_list)
 
         for pod in uninitialized_pods.values():
             _label_pod(namespace,

sky/provision/kubernetes/utils.py

@@ -1801,13 +1801,22 @@ def get_kubernetes_node_info(
     number of GPUs available on the node and the number of free GPUs on the
     node.
 
+    If the user does not have sufficient permissions to list pods in all
+    namespaces, the function will return free GPUs as -1.
+
     Returns:
         Dict[str, KubernetesNodeInfo]: Dictionary containing the node name as
             key and the KubernetesNodeInfo object as value
     """
     nodes = get_kubernetes_nodes(context)
     # Get the pods to get the real-time resource usage
-    pods = get_all_pods_in_kubernetes_cluster(context)
+    try:
+        pods = get_all_pods_in_kubernetes_cluster(context)
+    except kubernetes.api_exception() as e:
+        if e.status == 403:
+            pods = None
+        else:
+            raise
 
     label_formatter, _ = detect_gpu_label_formatter(context)
     if not label_formatter:
@@ -1828,19 +1837,22 @@ def get_kubernetes_node_info(
         accelerator_count = int(node.status.allocatable.get(
             'nvidia.com/gpu', 0))
 
-        for pod in pods:
-            # Get all the pods running on the node
-            if (pod.spec.node_name == node.metadata.name and
-                    pod.status.phase in ['Running', 'Pending']):
-                # Iterate over all the containers in the pod and sum the
-                # GPU requests
-                for container in pod.spec.containers:
-                    if container.resources.requests:
-                        allocated_qty += int(
-                            container.resources.requests.get(
-                                'nvidia.com/gpu', 0))
-
-        accelerators_available = accelerator_count - allocated_qty
+        if pods is None:
+            accelerators_available = -1
+
+        else:
+            for pod in pods:
+                # Get all the pods running on the node
+                if (pod.spec.node_name == node.metadata.name and
+                        pod.status.phase in ['Running', 'Pending']):
+                    # Iterate over all the containers in the pod and sum the
+                    # GPU requests
+                    for container in pod.spec.containers:
+                        if container.resources.requests:
+                            allocated_qty += int(
+                                container.resources.requests.get(
+                                    'nvidia.com/gpu', 0))
+            accelerators_available = accelerator_count - allocated_qty
 
         node_info_dict[node.metadata.name] = KubernetesNodeInfo(
             name=node.metadata.name,

sky/serve/__init__.py

@@ -11,6 +11,7 @@ from sky.serve.core import tail_logs
 from sky.serve.core import terminate_replica
 from sky.serve.core import up
 from sky.serve.core import update
+from sky.serve.load_balancing_policies import LB_POLICIES
 from sky.serve.serve_state import ReplicaStatus
 from sky.serve.serve_state import ServiceStatus
 from sky.serve.serve_utils import DEFAULT_UPDATE_MODE
@@ -35,6 +36,7 @@ __all__ = [
     'get_endpoint',
     'INITIAL_VERSION',
     'LB_CONTROLLER_SYNC_INTERVAL_SECONDS',
+    'LB_POLICIES',
     'ReplicaStatus',
     'ServiceComponent',
     'ServiceStatus',

sky/serve/load_balancer.py

@@ -2,7 +2,7 @@
 import asyncio
 import logging
 import threading
-from typing import Dict, Union
+from typing import Dict, Optional, Union
 
 import aiohttp
 import fastapi
@@ -27,18 +27,24 @@ class SkyServeLoadBalancer:
     policy.
     """
 
-    def __init__(self, controller_url: str, load_balancer_port: int) -> None:
+    def __init__(self,
+                 controller_url: str,
+                 load_balancer_port: int,
+                 load_balancing_policy_name: Optional[str] = None) -> None:
         """Initialize the load balancer.
 
         Args:
             controller_url: The URL of the controller.
             load_balancer_port: The port where the load balancer listens to.
+            load_balancing_policy_name: The name of the load balancing policy
+                to use. Defaults to None.
         """
         self._app = fastapi.FastAPI()
         self._controller_url: str = controller_url
         self._load_balancer_port: int = load_balancer_port
-        self._load_balancing_policy: lb_policies.LoadBalancingPolicy = (
-            lb_policies.RoundRobinPolicy())
+        # Use the registry to create the load balancing policy
+        self._load_balancing_policy = lb_policies.LoadBalancingPolicy.make(
+            load_balancing_policy_name)
         self._request_aggregator: serve_utils.RequestsAggregator = (
             serve_utils.RequestTimestamp())
         # TODO(tian): httpx.Client has a resource limit of 100 max connections
@@ -223,9 +229,21 @@ class SkyServeLoadBalancer:
         uvicorn.run(self._app, host='0.0.0.0', port=self._load_balancer_port)
 
 
-def run_load_balancer(controller_addr: str, load_balancer_port: int):
-    load_balancer = SkyServeLoadBalancer(controller_url=controller_addr,
-                                         load_balancer_port=load_balancer_port)
+def run_load_balancer(controller_addr: str,
+                      load_balancer_port: int,
+                      load_balancing_policy_name: Optional[str] = None) -> None:
+    """ Run the load balancer.
+
+    Args:
+        controller_addr: The address of the controller.
+        load_balancer_port: The port where the load balancer listens to.
+        policy_name: The name of the load balancing policy to use. Defaults to
+            None.
+    """
+    load_balancer = SkyServeLoadBalancer(
+        controller_url=controller_addr,
+        load_balancer_port=load_balancer_port,
+        load_balancing_policy_name=load_balancing_policy_name)
     load_balancer.run()
 
 
@@ -241,5 +259,13 @@ if __name__ == '__main__':
                         required=True,
                         default=8890,
                         help='The port where the load balancer listens to.')
+    available_policies = list(lb_policies.LB_POLICIES.keys())
+    parser.add_argument(
+        '--load-balancing-policy',
+        choices=available_policies,
+        default='round_robin',
+        help=f'The load balancing policy to use. Available policies: '
+        f'{", ".join(available_policies)}.')
     args = parser.parse_args()
-    run_load_balancer(args.controller_addr, args.load_balancer_port)
+    run_load_balancer(args.controller_addr, args.load_balancer_port,
+                      args.load_balancing_policy)

sky/serve/load_balancing_policies.py

@@ -10,6 +10,10 @@ if typing.TYPE_CHECKING:
 
 logger = sky_logging.init_logger(__name__)
 
+# Define a registry for load balancing policies
+LB_POLICIES = {}
+DEFAULT_LB_POLICY = None
+
 
 def _request_repr(request: 'fastapi.Request') -> str:
     return ('<Request '
@@ -25,6 +29,24 @@ class LoadBalancingPolicy:
     def __init__(self) -> None:
         self.ready_replicas: List[str] = []
 
+    def __init_subclass__(cls, name: str, default: bool = False):
+        LB_POLICIES[name] = cls
+        if default:
+            global DEFAULT_LB_POLICY
+            assert DEFAULT_LB_POLICY is None, (
+                'Only one policy can be default.')
+            DEFAULT_LB_POLICY = name
+
+    @classmethod
+    def make(cls, policy_name: Optional[str] = None) -> 'LoadBalancingPolicy':
+        """Create a load balancing policy from a name."""
+        if policy_name is None:
+            policy_name = DEFAULT_LB_POLICY
+
+        if policy_name not in LB_POLICIES:
+            raise ValueError(f'Unknown load balancing policy: {policy_name}')
+        return LB_POLICIES[policy_name]()
+
     def set_ready_replicas(self, ready_replicas: List[str]) -> None:
         raise NotImplementedError
 
@@ -44,7 +66,7 @@ class LoadBalancingPolicy:
         raise NotImplementedError
 
 
-class RoundRobinPolicy(LoadBalancingPolicy):
+class RoundRobinPolicy(LoadBalancingPolicy, name='round_robin', default=True):
     """Round-robin load balancing policy."""
 
     def __init__(self) -> None:

sky/serve/service.py

@@ -219,6 +219,9 @@ def _start(service_name: str, tmp_task_yaml: str, job_id: int):
             load_balancer_port = common_utils.find_free_port(
                 constants.LOAD_BALANCER_PORT_START)
 
+            # Extract the load balancing policy from the service spec
+            policy_name = service_spec.load_balancing_policy
+
             # Start the load balancer.
             # TODO(tian): Probably we could enable multiple ports specified in
             # service spec and we could start multiple load balancers.
@@ -227,7 +230,7 @@ def _start(service_name: str, tmp_task_yaml: str, job_id: int):
                 target=ux_utils.RedirectOutputForProcess(
                     load_balancer.run_load_balancer,
                     load_balancer_log_file).run,
-                args=(controller_addr, load_balancer_port))
+                args=(controller_addr, load_balancer_port, policy_name))
             load_balancer_process.start()
             serve_state.set_service_load_balancer_port(service_name,
                                                        load_balancer_port)

sky/serve/service_spec.py

@@ -6,6 +6,7 @@ from typing import Any, Dict, Optional
 
 import yaml
 
+from sky import serve
 from sky.serve import constants
 from sky.utils import common_utils
 from sky.utils import schemas
@@ -29,6 +30,7 @@ class SkyServiceSpec:
         base_ondemand_fallback_replicas: Optional[int] = None,
         upscale_delay_seconds: Optional[int] = None,
         downscale_delay_seconds: Optional[int] = None,
+        load_balancing_policy: Optional[str] = None,
     ) -> None:
         if max_replicas is not None and max_replicas < min_replicas:
             with ux_utils.print_exception_no_traceback():
@@ -55,6 +57,13 @@ class SkyServiceSpec:
                 raise ValueError('readiness_path must start with a slash (/). '
                                  f'Got: {readiness_path}')
 
+        # Add the check for unknown load balancing policies
+        if (load_balancing_policy is not None and
+                load_balancing_policy not in serve.LB_POLICIES):
+            with ux_utils.print_exception_no_traceback():
+                raise ValueError(
+                    f'Unknown load balancing policy: {load_balancing_policy}. '
+                    f'Available policies: {list(serve.LB_POLICIES.keys())}')
         self._readiness_path: str = readiness_path
         self._initial_delay_seconds: int = initial_delay_seconds
         self._readiness_timeout_seconds: int = readiness_timeout_seconds
@@ -69,6 +78,7 @@ class SkyServiceSpec:
             int] = base_ondemand_fallback_replicas
         self._upscale_delay_seconds: Optional[int] = upscale_delay_seconds
         self._downscale_delay_seconds: Optional[int] = downscale_delay_seconds
+        self._load_balancing_policy: Optional[str] = load_balancing_policy
 
         self._use_ondemand_fallback: bool = (
             self.dynamic_ondemand_fallback is not None and
@@ -150,6 +160,8 @@ class SkyServiceSpec:
             service_config['dynamic_ondemand_fallback'] = policy_section.get(
                 'dynamic_ondemand_fallback', None)
 
+        service_config['load_balancing_policy'] = config.get(
+            'load_balancing_policy', None)
         return SkyServiceSpec(**service_config)
 
     @staticmethod
@@ -205,6 +217,8 @@ class SkyServiceSpec:
                         self.upscale_delay_seconds)
         add_if_not_none('replica_policy', 'downscale_delay_seconds',
                         self.downscale_delay_seconds)
+        add_if_not_none('load_balancing_policy', None,
+                        self._load_balancing_policy)
         return config
 
     def probe_str(self):
@@ -256,6 +270,7 @@ class SkyServiceSpec:
             Readiness probe timeout seconds:  {self.readiness_timeout_seconds}
             Replica autoscaling policy:       {self.autoscaling_policy_str()}
             Spot Policy:                      {self.spot_policy_str()}
+            Load Balancing Policy:            {self.load_balancing_policy}
         """)
 
     @property
@@ -310,3 +325,7 @@ class SkyServiceSpec:
     @property
     def use_ondemand_fallback(self) -> bool:
         return self._use_ondemand_fallback
+
+    @property
+    def load_balancing_policy(self) -> Optional[str]:
+        return self._load_balancing_policy

sky/templates/kubernetes-ray.yml.j2

@@ -324,6 +324,15 @@ available_node_types:
           command: ["/bin/bash", "-c", "--"]
           args: 
             - |
+              # Helper function to conditionally use sudo
+              prefix_cmd() { if [ $(id -u) -ne 0 ]; then echo "sudo"; else echo ""; fi; }
+
+              # Run apt update in background and log to a file
+              (
+                DEBIAN_FRONTEND=noninteractive $(prefix_cmd) apt-get update > /tmp/apt-update.log 2>&1 || \
+                echo "Warning: apt-get update failed. Continuing anyway..." >> /tmp/apt-update.log
+              ) &
+
               function mylsof { p=$(for pid in /proc/{0..9}*; do i=$(basename "$pid"); for file in "$pid"/fd/*; do link=$(readlink -e "$file"); if [ "$link" = "$1" ]; then echo "$i"; fi; done; done); echo "$p"; };
 
               # Tails file and checks every 5 sec for
@@ -419,7 +428,18 @@ setup_commands:
   # Line 'sudo grep ..': set the number of threads per process to unlimited to avoid ray job submit stucking issue when the number of running ray jobs increase.
   # Line 'mkdir -p ..': disable host key check
   # Line 'python3 -c ..': patch the buggy ray files and enable `-o allow_other` option for `goofys`
-  - sudo DEBIAN_FRONTEND=noninteractive apt install lsof gcc patch pciutils rsync fuse curl -y;
+  - |
+    PACKAGES="gcc patch pciutils rsync fuse curl";
+    MISSING_PACKAGES="";
+    for pkg in $PACKAGES; do
+      if ! dpkg -l | grep -q "^ii  $pkg "; then
+        MISSING_PACKAGES="$MISSING_PACKAGES $pkg";
+      fi
+    done;
+    if [ ! -z "$MISSING_PACKAGES" ]; then
+      echo "Installing missing packages: $MISSING_PACKAGES";
+      sudo DEBIAN_FRONTEND=noninteractive apt-get install -y $MISSING_PACKAGES;
+    fi;
     mkdir -p ~/.ssh; touch ~/.ssh/config;
     {%- for initial_setup_command in initial_setup_commands %}
     {{ initial_setup_command }}

sky/utils/kubernetes/generate_kubeconfig.sh

@@ -112,6 +112,9 @@ rules:
   - apiGroups: ["networking.k8s.io"]   # Required for exposing services through ingresses
     resources: ["ingressclasses"]
     verbs: ["get", "list", "watch"]
+  - apiGroups: [""]                 # Required for sky show-gpus command
+    resources: ["pods"]
+    verbs: ["get", "list"]
 ---
 # ClusterRoleBinding for the service account
 apiVersion: rbac.authorization.k8s.io/v1

sky/utils/schemas.py

@@ -308,6 +308,9 @@ def get_storage_schema():
 
 def get_service_schema():
     """Schema for top-level `service:` field (for SkyServe)."""
+    # To avoid circular imports, only import when needed.
+    # pylint: disable=import-outside-toplevel
+    from sky.serve import load_balancing_policies
     return {
         '$schema': 'https://json-schema.org/draft/2020-12/schema',
         'type': 'object',
@@ -382,6 +385,11 @@ def get_service_schema():
             'replicas': {
                 'type': 'integer',
             },
+            'load_balancing_policy': {
+                'type': 'string',
+                'case_insensitive_enum': list(
+                    load_balancing_policies.LB_POLICIES.keys())
+            },
         }
     }
 

{skypilot_nightly-1.0.0.dev20241110.dist-info → skypilot_nightly-1.0.0.dev20241112.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: skypilot-nightly
-Version: 1.0.0.dev20241110
+Version: 1.0.0.dev20241112
 Summary: SkyPilot: An intercloud broker for the clouds
 Author: SkyPilot Team
 License: Apache 2.0
@@ -309,7 +309,7 @@ Runnable examples:
   - [LocalGPT](./llm/localgpt)
   - [Falcon](./llm/falcon)
   - Add yours here & see more in [`llm/`](./llm)!
-- Framework examples: [PyTorch DDP](https://github.com/skypilot-org/skypilot/blob/master/examples/resnet_distributed_torch.yaml), [DeepSpeed](./examples/deepspeed-multinode/sky.yaml), [JAX/Flax on TPU](https://github.com/skypilot-org/skypilot/blob/master/examples/tpu/tpuvm_mnist.yaml), [Stable Diffusion](https://github.com/skypilot-org/skypilot/tree/master/examples/stable_diffusion), [Detectron2](https://github.com/skypilot-org/skypilot/blob/master/examples/detectron2_docker.yaml), [Distributed](https://github.com/skypilot-org/skypilot/blob/master/examples/resnet_distributed_tf_app.py) [TensorFlow](https://github.com/skypilot-org/skypilot/blob/master/examples/resnet_app_storage.yaml), [Ray Train](examples/distributed_ray_train/ray_train.yaml), [NeMo](https://github.com/skypilot-org/skypilot/blob/master/examples/nemo/nemo.yaml), [programmatic grid search](https://github.com/skypilot-org/skypilot/blob/master/examples/huggingface_glue_imdb_grid_search_app.py), [Docker](https://github.com/skypilot-org/skypilot/blob/master/examples/docker/echo_app.yaml), [Cog](https://github.com/skypilot-org/skypilot/blob/master/examples/cog/), [Unsloth](https://github.com/skypilot-org/skypilot/blob/master/examples/unsloth/unsloth.yaml), [Ollama](https://github.com/skypilot-org/skypilot/blob/master/llm/ollama), [llm.c](https://github.com/skypilot-org/skypilot/tree/master/llm/gpt-2), [Airflow](./examples/airflow/training_workflow) and [many more (`examples/`)](./examples).
+- Framework examples: [PyTorch DDP](https://github.com/skypilot-org/skypilot/blob/master/examples/resnet_distributed_torch.yaml), [DeepSpeed](./examples/deepspeed-multinode/sky.yaml), [JAX/Flax on TPU](https://github.com/skypilot-org/skypilot/blob/master/examples/tpu/tpuvm_mnist.yaml), [Stable Diffusion](https://github.com/skypilot-org/skypilot/tree/master/examples/stable_diffusion), [Detectron2](https://github.com/skypilot-org/skypilot/blob/master/examples/detectron2_docker.yaml), [Distributed](https://github.com/skypilot-org/skypilot/blob/master/examples/resnet_distributed_tf_app.py) [TensorFlow](https://github.com/skypilot-org/skypilot/blob/master/examples/resnet_app_storage.yaml), [Ray Train](examples/distributed_ray_train/ray_train.yaml), [NeMo](https://github.com/skypilot-org/skypilot/blob/master/examples/nemo/), [programmatic grid search](https://github.com/skypilot-org/skypilot/blob/master/examples/huggingface_glue_imdb_grid_search_app.py), [Docker](https://github.com/skypilot-org/skypilot/blob/master/examples/docker/echo_app.yaml), [Cog](https://github.com/skypilot-org/skypilot/blob/master/examples/cog/), [Unsloth](https://github.com/skypilot-org/skypilot/blob/master/examples/unsloth/unsloth.yaml), [Ollama](https://github.com/skypilot-org/skypilot/blob/master/llm/ollama), [llm.c](https://github.com/skypilot-org/skypilot/tree/master/llm/gpt-2), [Airflow](./examples/airflow/training_workflow) and [many more (`examples/`)](./examples).
 
 Case Studies and Integrations: [Community Spotlights](https://blog.skypilot.co/community/)
 

{skypilot_nightly-1.0.0.dev20241110.dist-info → skypilot_nightly-1.0.0.dev20241112.dist-info}/RECORD

@@ -1,8 +1,8 @@
-sky/__init__.py,sha256=j3vy9X4XOYIefQk15d_c6Q_mpDjII9Nltso4xgrFI1o,5882
+sky/__init__.py,sha256=8VuuTyDTVZB1BaeWD7OwBFZwpwweQkb0DNyPpsitRQs,5882
 sky/admin_policy.py,sha256=hPo02f_A32gCqhUueF0QYy1fMSSKqRwYEg_9FxScN_s,3248
 sky/authentication.py,sha256=pAdCT60OxxiXI9KXDyP2lQ9u9vMc6aMtq5Xi2h_hbdw,20984
 sky/check.py,sha256=D3Y3saIFAYVvPxuBHnVgJEO0fUVDxgjwuMBaO-D778k,9472
-sky/cli.py,sha256=jEjXs5Z0u263eJIsTHoKyG9oOY6giqw19s2di9kEv1s,212088
+sky/cli.py,sha256=oGBQrCYVWqRTcWR-yCKZY7dmUOUnP5Xuvz_zcFXzqlw,212342
 sky/cloud_stores.py,sha256=RjFgmRhUh1Kk__f6g3KxzLp9s7dA0pFK4W1AukEuUaw,21153
 sky/core.py,sha256=0-4W_DKJZgbwXuzNZKQ2R_qJxqxbqqNfyi0U0PQBKvQ,38230
 sky/dag.py,sha256=f3sJlkH4bE6Uuz3ozNtsMhcBpRx7KmC9Sa4seDKt4hU,3104
@@ -65,7 +65,7 @@ sky/clouds/service_catalog/cudo_catalog.py,sha256=V_takvL6dWTGQaTLCEvjKIotCDPnMu
 sky/clouds/service_catalog/fluidstack_catalog.py,sha256=21-cvrYEYTIi7n3ZNF2e7_0QX-PF4BkhlVJUWQOvKrY,5059
 sky/clouds/service_catalog/gcp_catalog.py,sha256=v_5fsB3dB9oD8U7lBKnCe5ii6AUWEOiQjNarMnU_qLA,24379
 sky/clouds/service_catalog/ibm_catalog.py,sha256=1iK0KvbI82U7sySb7chr-qm_16x3tTnZ6nIo7o76ouc,4493
-sky/clouds/service_catalog/kubernetes_catalog.py,sha256=5ilQ-JK1ZS2EZp8GpCKok0H3S1fdI_aAznzIDWCY1NY,9110
+sky/clouds/service_catalog/kubernetes_catalog.py,sha256=c6Oot8RC1ujcFmfJbkeJKUWsw3aX0iNvKL1fJg-FoOc,10020
 sky/clouds/service_catalog/lambda_catalog.py,sha256=2R-ccu63BbdvO6X80MtxiniA-jLewXb6I0Ye1rYD9fY,5302
 sky/clouds/service_catalog/oci_catalog.py,sha256=cyA6ZqwHGOKuPxUl_dKmFGdeWdQGMrvl_-o2MtyF998,8580
 sky/clouds/service_catalog/paperspace_catalog.py,sha256=MOlfoGRChjEwMzu4nRAho8DrIwwUJ3QlRzrMA1RLqvE,3789
@@ -106,7 +106,7 @@ sky/jobs/dashboard/templates/index.html,sha256=su1tqgcsXNl1lGl9hfIR6ig1f531OO57x
 sky/provision/__init__.py,sha256=llAtnAAzx0TKT17B0JL_2ZiKea9RRQRxSzkWHQYqWTo,6292
 sky/provision/common.py,sha256=E8AlSUFcn0FYQq1erNmoVfMAdsF9tP2yxfyk-9PLvQU,10286
 sky/provision/constants.py,sha256=oc_XDUkcoLQ_lwDy5yMeMSWviKS0j0s1c0pjlvpNeWY,800
-sky/provision/docker_utils.py,sha256=cKYasCwbMf6C2_0vTxg2GvbrnhFvko-xDl1frfm7wxc,19199
+sky/provision/docker_utils.py,sha256=l4AMzwXGZd8RyNq8AwOaKV9bFSofLYfSyj2NBhkXYsY,19200
 sky/provision/instance_setup.py,sha256=gI739UMCqtPqdA522D92bPu5sA3OHBMDmIGmqqxsIwY,23652
 sky/provision/logging.py,sha256=yZWgejrFBhhRjAtvFu5N5bRXIMK5TuwNjp1vKQqz2pw,2103
 sky/provision/metadata_utils.py,sha256=LrxeV4wD2QPzNdXV_npj8q-pr35FatxBBjF_jSbpOT0,4013
@@ -137,10 +137,10 @@ sky/provision/gcp/instance_utils.py,sha256=veRBr6Oziv0KaUdC4acuWeaOremNV0gMYCCHa
 sky/provision/gcp/mig_utils.py,sha256=oFpcFZoapHMILSE4iIm8V5bxP1RhbMHRF7cciqq8qAk,7883
 sky/provision/kubernetes/__init__.py,sha256=y6yVfii81WYG3ROxv4hiIj-ydinS5-xGxLvXnARVQoI,719
 sky/provision/kubernetes/config.py,sha256=WEKcFXXhe89bLGAvoMiBvTDxdxkpTIA6ezrj2vmzldc,29072
-sky/provision/kubernetes/instance.py,sha256=rY43hZOInP20kYofW0MGs7wDbJ4NxMw1FtKAJAPGIOU,43960
+sky/provision/kubernetes/instance.py,sha256=MFtTh-dNIuTZcHD20PQG_QuULFRFaPxwlUczR6sRnsk,43601
 sky/provision/kubernetes/network.py,sha256=EpNjRQ131CXepqbdkoRKFu4szVrm0oKEpv1l8EgOkjU,12364
 sky/provision/kubernetes/network_utils.py,sha256=t1FS3K400fetH7cBuRgQJZl5_jEeMshsvsYmnMUcq8k,11399
-sky/provision/kubernetes/utils.py,sha256=2N5c4yA7CEn4DjvCiUO73W4XDEjgixcJRVdgs913QQE,89523
+sky/provision/kubernetes/utils.py,sha256=PEDyZnf-dSmQ4dXyS_0x9OYHt9SbY7A6urd436f-WyQ,89923
 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/lambda_cloud/__init__.py,sha256=6EEvSgtUeEiup9ivIFevHmgv0GqleroO2X0K7TRa2nE,612
@@ -175,18 +175,18 @@ sky/provision/vsphere/common/service_manager_factory.py,sha256=YkvfHiRXFK_Nb406z
 sky/provision/vsphere/common/ssl_helper.py,sha256=TYzN9K0i_Mk_17PKGyGPgvOGfoizysuuIeYapcy_tWE,795
 sky/provision/vsphere/common/vapiconnect.py,sha256=R2I1ZWBA19d11fZ_FrIzQT8E1aLl1HU4Rdcj8Z5r3NE,2932
 sky/provision/vsphere/common/vim_utils.py,sha256=EMWLS8ILpdx6XwUZ9I53y0B_1yFrRrlr4jjIMT84hAc,17877
-sky/serve/__init__.py,sha256=gFZt7W3UPMi4qvYe2xgkHg1VxbR1WGavKyWLBUD3mpg,1731
+sky/serve/__init__.py,sha256=Bqw8nB9u1QF3ryjbV797SPZq0DWAcjT94E_5B8J24ag,1808
 sky/serve/autoscalers.py,sha256=khY1oZ22PRaUQNsLCoNKH178X_NiJw0LSLOKr7_LNgY,30275
 sky/serve/constants.py,sha256=7MflfgTHO9gDSux93U4BmNeEMWXxZB4q7I54KUwgp-s,4651
 sky/serve/controller.py,sha256=R5iIEGEEFtbm_6MvSGelYZP-vSmW0cSFuy64OexUc4g,11719
 sky/serve/core.py,sha256=hszs95BwtC4wIJujGNokvFC46VjojgRz1BbYOIIPh6k,31601
-sky/serve/load_balancer.py,sha256=aUfDsgUT_fYrchCwJCeunMPXmAkwJAY58BEu-IN2FaA,11571
-sky/serve/load_balancing_policies.py,sha256=ExdwH_pxPYpJ6CkoTQCOPSa4lzwbq1LFFMKzmIu8ryk,2331
+sky/serve/load_balancer.py,sha256=I4W66eh1t1kA_C_VaMPI76WeDTCl3Z6rFxF6rQIWd6E,12636
+sky/serve/load_balancing_policies.py,sha256=_k4tkwIvhulR02Ln9ixYB_b97KOypr2xfSjMx8_zky0,3143
 sky/serve/replica_managers.py,sha256=1xYDK9Te5wFEF5hUK0gyNIUib0MY-HScLHUBDlTSl-k,57774
 sky/serve/serve_state.py,sha256=Q7De4GoBEPxlN_t1Lpn-Y1fd94SeHZ3E-94f1OTuhpc,19086
 sky/serve/serve_utils.py,sha256=wqBxChpJylZ_qHWyFmMBJqrG8_7xTIOr9nlOeyHs9P8,39431
-sky/serve/service.py,sha256=fkfJvNJ2BO6rfV0TblZG-QkOXaCyZlpkwbGgrsTzf2w,11872
-sky/serve/service_spec.py,sha256=1aS6b-ku7W4CjyekXKDxjZsDdt-O8ygos-jFeXu31cA,13766
+sky/serve/service.py,sha256=gVem2vX8XuR_1wTqwrzbszQAbjzjDP2ddd787aynT9g,12017
+sky/serve/service_spec.py,sha256=34dMQ37INHltBzWaxHl3y_o9X3wLOCWA5jUhmhH1II4,14740
 sky/setup_files/MANIFEST.in,sha256=WF0T89NLichHxZDDSQzvSpiONtAEFyur2MPmGczgTIo,555
 sky/setup_files/setup.py,sha256=G767GNB-jXqyC8MR-IdiojnnI2E6tP4gMYenKU14ZGA,12156
 sky/skylet/LICENSE,sha256=BnFrJSvUFpMUoH5mOpWnEvaC5R6Uux8W6WXgrte8iYg,12381
@@ -228,7 +228,7 @@ sky/templates/jobs-controller.yaml.j2,sha256=Gu3ogFxFYr09VEXP-6zEbrCUOFo1aYxWEjA
 sky/templates/kubernetes-ingress.yml.j2,sha256=73iDklVDWBMbItg0IexCa6_ClXPJOxw7PWz3leku4nE,1340
 sky/templates/kubernetes-loadbalancer.yml.j2,sha256=IxrNYM366N01bbkJEbZ_UPYxUP8wyVEbRNFHRsBuLsw,626
 sky/templates/kubernetes-port-forward-proxy-command.sh,sha256=HlG7CPBBedCVBlL9qv0erW_eKm6Irj0LFyaAWuJW_lc,3148
-sky/templates/kubernetes-ray.yml.j2,sha256=dsWlkX-0b1igeZI4c0u0Jzia5I_9gezCiewR6pX1LlY,18374
+sky/templates/kubernetes-ray.yml.j2,sha256=Ek6nePe_IP1b0mqMLnbyjp7wpo1-kwranD_AFRXJ9tU,19152
 sky/templates/kubernetes-ssh-jump.yml.j2,sha256=k5W5sOIMppU7dDkJMwPlqsUcb92y7L5_TVG3hkgMy8M,2747
 sky/templates/lambda-ray.yml.j2,sha256=HyvO_tX2vxwSsc4IFVSqGuIbjLMk0bevP9bcxb8ZQII,4498
 sky/templates/local-ray.yml.j2,sha256=FNHeyHF6nW9nU9QLIZceUWfvrFTTcO51KqhTnYCEFaA,1185
@@ -257,7 +257,7 @@ sky/utils/kubernetes_enums.py,sha256=imGqHSa8O07zD_6xH1SDMM7dBU5lF5fzFFlQuQy00QM
 sky/utils/log_utils.py,sha256=ptv2sbsiJSgk4NvdccrMsUR-MvOKnbu4BQiRSishgk0,12472
 sky/utils/resources_utils.py,sha256=Xqi7gxPYw2y5wl5okUI5zx5LEij0hJF_V3Zi8q7TXYg,7890
 sky/utils/rich_utils.py,sha256=hmnI1X5dKvRIQzB7EyNb34FT97qFNve-0QHqM5r0mVk,3066
-sky/utils/schemas.py,sha256=mogoStpQ77S936VfChinAW2I1DT4q2c5E7qY_qNiO0w,29094
+sky/utils/schemas.py,sha256=67LK87wBywblIyF-QgG5hgL1BvBuHsxeQLQBO0M5OH4,29447
 sky/utils/subprocess_utils.py,sha256=mMFCTfxbyav5LJ1epJJXkgfFYmd828naTOMVfYjuEWY,6905
 sky/utils/timeline.py,sha256=ao_nm0y52ZQILfL7Y92c3pSEFRyPm_ElORC3DrI5BwQ,3936
 sky/utils/ux_utils.py,sha256=CqyIFGDuSE8fQasPkna_loZMwtboC9KedR09WEQ7qz0,6502
@@ -269,15 +269,15 @@ sky/utils/kubernetes/create_cluster.sh,sha256=VLXfazav9XCMQmeKVqhuOQzt2vM6G1jgnv
 sky/utils/kubernetes/delete_cluster.sh,sha256=BSccHF43GyepDNf-FZcenzHzpXXATkVD92vgn1lWPgk,927
 sky/utils/kubernetes/deploy_remote_cluster.sh,sha256=vGj0mD0tejHDRy8ulwKOvOF2mfLyT5J8fp7GVqEe_EY,8478
 sky/utils/kubernetes/generate_kind_config.py,sha256=_TNLnifA_r7-CRq083IP1xjelYqiLjzQX9ohuqYpDH8,3187
-sky/utils/kubernetes/generate_kubeconfig.sh,sha256=AcYhuuG5jXWGHUmyRuH-oKy5qcn92gXhu6bXOt6eD6g,9274
+sky/utils/kubernetes/generate_kubeconfig.sh,sha256=livvxDKV-_xx8-dYWNyo4wlg3sOldeHefI37JXKLXu0,9398
 sky/utils/kubernetes/gpu_labeler.py,sha256=MEUv0U4ACDcNwtFVltlv017XJMjxx1Bndf6fL0i6eqg,6960
 sky/utils/kubernetes/k8s_gpu_labeler_job.yaml,sha256=k0TBoQ4zgf79-sVkixKSGYFHQ7ZWF5gdVIZPupCCo9A,1224
 sky/utils/kubernetes/k8s_gpu_labeler_setup.yaml,sha256=VLKT2KKimZu1GDg_4AIlIt488oMQvhRZWwsj9vBbPUg,3812
 sky/utils/kubernetes/rsync_helper.sh,sha256=hyYDaYSNxYaNvzUQBzC8AidB7nDeojizjkzc_CTxycY,1077
 sky/utils/kubernetes/ssh_jump_lifecycle_manager.py,sha256=RFLJ3k7MR5UN4SKHykQ0lV9SgXumoULpKYIAt1vh-HU,6560
-skypilot_nightly-1.0.0.dev20241110.dist-info/LICENSE,sha256=emRJAvE7ngL6x0RhQvlns5wJzGI3NEQ_WMjNmd9TZc4,12170
-skypilot_nightly-1.0.0.dev20241110.dist-info/METADATA,sha256=4ar4pUczmGqsEHMG-85ANcAB_ifYgIDJRr0BJfypruA,19708
-skypilot_nightly-1.0.0.dev20241110.dist-info/WHEEL,sha256=P9jw-gEje8ByB7_hXoICnHtVCrEwMQh-630tKvQWehc,91
-skypilot_nightly-1.0.0.dev20241110.dist-info/entry_points.txt,sha256=StA6HYpuHj-Y61L2Ze-hK2IcLWgLZcML5gJu8cs6nU4,36
-skypilot_nightly-1.0.0.dev20241110.dist-info/top_level.txt,sha256=qA8QuiNNb6Y1OF-pCUtPEr6sLEwy2xJX06Bd_CrtrHY,4
-skypilot_nightly-1.0.0.dev20241110.dist-info/RECORD,,
+skypilot_nightly-1.0.0.dev20241112.dist-info/LICENSE,sha256=emRJAvE7ngL6x0RhQvlns5wJzGI3NEQ_WMjNmd9TZc4,12170
+skypilot_nightly-1.0.0.dev20241112.dist-info/METADATA,sha256=Ui6L9CmuvZsIg2D0paU-NiqfVLtywzq5GLpCrJes-eY,19699
+skypilot_nightly-1.0.0.dev20241112.dist-info/WHEEL,sha256=a7TGlA-5DaHMRrarXjVbQagU3Man_dCnGIWMJr5kRWo,91
+skypilot_nightly-1.0.0.dev20241112.dist-info/entry_points.txt,sha256=StA6HYpuHj-Y61L2Ze-hK2IcLWgLZcML5gJu8cs6nU4,36
+skypilot_nightly-1.0.0.dev20241112.dist-info/top_level.txt,sha256=qA8QuiNNb6Y1OF-pCUtPEr6sLEwy2xJX06Bd_CrtrHY,4
+skypilot_nightly-1.0.0.dev20241112.dist-info/RECORD,,

{skypilot_nightly-1.0.0.dev20241110.dist-info → skypilot_nightly-1.0.0.dev20241112.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (75.3.0)
+Generator: setuptools (75.4.0)
 Root-Is-Purelib: true
 Tag: py3-none-any