skypilot-nightly 1.0.0.dev20250129__py3-none-any.whl → 1.0.0.dev20250130__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 = 'a80208ff0f4c4c8ce869b1a693e5d8d9e588ba52'
+_SKYPILOT_COMMIT_SHA = 'db90a41f4dff842aaaacd105d77f9e9a4e29b4bc'
 
 
 def _get_git_commit():
@@ -35,7 +35,7 @@ def _get_git_commit():
 
 
 __commit__ = _get_git_commit()
-__version__ = '1.0.0.dev20250129'
+__version__ = '1.0.0.dev20250130'
 __root_dir__ = os.path.dirname(os.path.abspath(__file__))
 
 

sky/cli.py

@@ -4059,32 +4059,64 @@ def _generate_task_with_service(
         with ux_utils.print_exception_no_traceback():
             raise ValueError('Service section not found in the YAML file. '
                              'To fix, add a valid `service` field.')
-    service_port: Optional[int] = None
-    for requested_resources in list(task.resources):
-        if requested_resources.ports is None or len(
-                requested_resources.ports) != 1:
-            with ux_utils.print_exception_no_traceback():
-                raise ValueError(
-                    'Must only specify one port in resources. Each replica '
-                    'will use the port specified as application ingress port.')
-        service_port_str = requested_resources.ports[0]
-        if not service_port_str.isdigit():
-            # For the case when the user specified a port range like 10000-10010
-            with ux_utils.print_exception_no_traceback():
-                raise ValueError(f'Port {service_port_str!r} is not a valid '
-                                 'port number. Please specify a single port '
-                                 f'instead. Got: {service_port_str!r}')
-        # We request all the replicas using the same port for now, but it
-        # should be fine to allow different replicas to use different ports
-        # in the future.
-        resource_port = int(service_port_str)
-        if service_port is None:
-            service_port = resource_port
-        if service_port != resource_port:
-            with ux_utils.print_exception_no_traceback():
-                raise ValueError(f'Got multiple ports: {service_port} and '
-                                 f'{resource_port} in different resources. '
-                                 'Please specify single port instead.')
+
+    # NOTE(yi): we only allow one service port now.
+    service_port: Optional[int] = int(
+        task.service.ports) if task.service.ports is not None else None
+    if service_port is None:
+        for requested_resources in list(task.resources):
+            if requested_resources.ports is None:
+                with ux_utils.print_exception_no_traceback():
+                    raise ValueError(
+                        'Must specify at least one ports in resources. Each '
+                        'replica will use the port specified as application '
+                        'ingress port if only one port is specified in the '
+                        'replica resources. If there are multiple ports opened '
+                        'in the replica, please set the `service.ports` field '
+                        'in the service config.')
+            requested_ports = list(
+                resources_utils.port_ranges_to_set(requested_resources.ports))
+            if len(requested_ports) > 1:
+                with ux_utils.print_exception_no_traceback():
+                    raise ValueError(
+                        'Multiple ports specified in resources. Please '
+                        'specify the main port in the `service.ports` field.')
+            # We request all the replicas using the same port for now, but it
+            # should be fine to allow different replicas to use different ports
+            # in the future.
+            resource_port = requested_ports[0]
+            if service_port is None:
+                service_port = resource_port
+            if service_port != resource_port:
+                with ux_utils.print_exception_no_traceback():
+                    raise ValueError(
+                        f'Got multiple ports: {service_port} and '
+                        f'{resource_port} in different resources. '
+                        'Please specify the same port in all replicas, or '
+                        'explicitly set the service port in the '
+                        '`service.ports` section.')
+        assert service_port is not None
+        task.service.set_ports(str(service_port))
+    else:
+        for requested_resources in list(task.resources):
+            if requested_resources.ports is None:
+                with ux_utils.print_exception_no_traceback():
+                    raise ValueError(
+                        'Must specify at least one ports in every replica '
+                        'resources.')
+            ports_set = resources_utils.port_ranges_to_set(
+                requested_resources.ports)
+            if service_port not in ports_set:
+                with ux_utils.print_exception_no_traceback():
+                    # TODO(tian): Automatically infer resource port from
+                    # service port if none of them is specified in the
+                    # replica resources.
+                    raise ValueError(
+                        f'The service port {service_port} specified in the '
+                        'service section is not found in some resources. '
+                        'Please check if the service port is correct or add '
+                        'the service port to replica resources.')
+
     return task
 
 

sky/jobs/core.py

@@ -2,7 +2,7 @@
 import os
 import tempfile
 import typing
-from typing import Any, Dict, List, Optional, Union
+from typing import Any, Dict, List, Optional, Tuple, Union
 import uuid
 
 import colorama
@@ -37,13 +37,13 @@ if typing.TYPE_CHECKING:
 @timeline.event
 @usage_lib.entrypoint
 def launch(
-        task: Union['sky.Task', 'sky.Dag'],
-        name: Optional[str] = None,
-        stream_logs: bool = True,
-        detach_run: bool = False,
-        # TODO(cooperc): remove fast arg before 0.8.0
-        fast: bool = True,  # pylint: disable=unused-argument for compatibility
-) -> None:
+    task: Union['sky.Task', 'sky.Dag'],
+    name: Optional[str] = None,
+    stream_logs: bool = True,
+    detach_run: bool = False,
+    # TODO(cooperc): remove fast arg before 0.8.0
+    fast: bool = True,  # pylint: disable=unused-argument for compatibility
+) -> Tuple[Optional[int], Optional[backends.ResourceHandle]]:
     # NOTE(dev): Keep the docstring consistent between the Python API and CLI.
     """Launch a managed job.
 
@@ -61,6 +61,13 @@ def launch(
         ValueError: cluster does not exist. Or, the entrypoint is not a valid
             chain dag.
         sky.exceptions.NotSupportedError: the feature is not supported.
+
+    Returns:
+      job_id: Optional[int]; the job ID of the submitted job. None if the
+        backend is not CloudVmRayBackend, or no job is submitted to
+        the cluster.
+      handle: Optional[backends.ResourceHandle]; handle to the controller VM.
+        None if dryrun.
     """
     entrypoint = task
     dag_uuid = str(uuid.uuid4().hex[:4])
@@ -140,15 +147,15 @@ def launch(
             f'{colorama.Fore.YELLOW}'
             f'Launching managed job {dag.name!r} from jobs controller...'
             f'{colorama.Style.RESET_ALL}')
-        sky.launch(task=controller_task,
-                   stream_logs=stream_logs,
-                   cluster_name=controller_name,
-                   detach_run=detach_run,
-                   idle_minutes_to_autostop=skylet_constants.
-                   CONTROLLER_IDLE_MINUTES_TO_AUTOSTOP,
-                   retry_until_up=True,
-                   fast=True,
-                   _disable_controller_check=True)
+        return sky.launch(task=controller_task,
+                          stream_logs=stream_logs,
+                          cluster_name=controller_name,
+                          detach_run=detach_run,
+                          idle_minutes_to_autostop=skylet_constants.
+                          CONTROLLER_IDLE_MINUTES_TO_AUTOSTOP,
+                          retry_until_up=True,
+                          fast=True,
+                          _disable_controller_check=True)
 
 
 def queue_from_kubernetes_pod(

sky/serve/core.py

@@ -68,7 +68,8 @@ def _validate_service_task(task: 'sky.Task') -> None:
                                  'SkyServe will replenish preempted spot '
                                  f'with {policy_description} instances.')
 
-    replica_ingress_port: Optional[int] = None
+    replica_ingress_port: Optional[int] = int(
+        task.service.ports) if (task.service.ports is not None) else None
     for requested_resources in task.resources:
         if (task.service.use_ondemand_fallback and
                 not requested_resources.use_spot):
@@ -77,22 +78,26 @@ def _validate_service_task(task: 'sky.Task') -> None:
                     '`use_ondemand_fallback` is only supported '
                     'for spot resources. Please explicitly specify '
                     '`use_spot: true` in resources for on-demand fallback.')
-        requested_ports = list(
-            resources_utils.port_ranges_to_set(requested_resources.ports))
-        if len(requested_ports) != 1:
-            with ux_utils.print_exception_no_traceback():
-                raise ValueError(
-                    'Must only specify one port in resources. Each replica '
-                    'will use the port specified as application ingress port.')
-        service_port = requested_ports[0]
-        if replica_ingress_port is None:
-            replica_ingress_port = service_port
-        elif service_port != replica_ingress_port:
-            with ux_utils.print_exception_no_traceback():
-                raise ValueError(
-                    f'Got multiple ports: {service_port} and '
-                    f'{replica_ingress_port} in different resources. '
-                    'Please specify the same port instead.')
+        if task.service.ports is None:
+            requested_ports = list(
+                resources_utils.port_ranges_to_set(requested_resources.ports))
+            if len(requested_ports) != 1:
+                with ux_utils.print_exception_no_traceback():
+                    raise ValueError(
+                        'To open multiple ports on the replica, please set the '
+                        '`service.ports` field to specify a main service port. '
+                        'Must only specify one port in resources otherwise. '
+                        'Each replica will use the port specified as '
+                        'application ingress port.')
+            service_port = requested_ports[0]
+            if replica_ingress_port is None:
+                replica_ingress_port = service_port
+            elif service_port != replica_ingress_port:
+                with ux_utils.print_exception_no_traceback():
+                    raise ValueError(
+                        f'Got multiple ports: {service_port} and '
+                        f'{replica_ingress_port} in different resources. '
+                        'Please specify the same port instead.')
 
 
 def _rewrite_tls_credential_paths_and_get_tls_env_vars(

sky/serve/replica_managers.py

@@ -171,13 +171,11 @@ def terminate_cluster(cluster_name: str,
 def _get_resources_ports(task_yaml: str) -> str:
     """Get the resources ports used by the task."""
     task = sky.Task.from_yaml(task_yaml)
-    # Already checked all ports are the same in sky.serve.core.up
+    # Already checked all ports are valid in sky.serve.core.up
     assert task.resources, task
-    task_resources: 'resources.Resources' = list(task.resources)[0]
-    # Already checked the resources have and only have one port
-    # before upload the task yaml.
-    assert task_resources.ports is not None
-    return task_resources.ports[0]
+    assert task.service is not None, task
+    assert task.service.ports is not None, task
+    return task.service.ports
 
 
 def _should_use_spot(task_yaml: str,

sky/serve/service_spec.py

@@ -25,6 +25,7 @@ class SkyServiceSpec:
         readiness_timeout_seconds: int,
         min_replicas: int,
         max_replicas: Optional[int] = None,
+        ports: Optional[str] = None,
         target_qps_per_replica: Optional[float] = None,
         post_data: Optional[Dict[str, Any]] = None,
         tls_credential: Optional[serve_utils.TLSCredential] = None,
@@ -72,6 +73,7 @@ class SkyServiceSpec:
         self._readiness_timeout_seconds: int = readiness_timeout_seconds
         self._min_replicas: int = min_replicas
         self._max_replicas: Optional[int] = max_replicas
+        self._ports: Optional[str] = ports
         self._target_qps_per_replica: Optional[float] = target_qps_per_replica
         self._post_data: Optional[Dict[str, Any]] = post_data
         self._tls_credential: Optional[serve_utils.TLSCredential] = (
@@ -137,6 +139,14 @@ class SkyServiceSpec:
         service_config['post_data'] = post_data
         service_config['readiness_headers'] = readiness_headers
 
+        ports = config.get('ports', None)
+        if ports is not None:
+            assert isinstance(ports, int)
+            if not 1 <= ports <= 65535:
+                with ux_utils.print_exception_no_traceback():
+                    raise ValueError('Port must be between 1 and 65535.')
+        service_config['ports'] = str(ports) if ports is not None else None
+
         policy_section = config.get('replica_policy', None)
         simplified_policy_section = config.get('replicas', None)
         if policy_section is None or simplified_policy_section is not None:
@@ -235,6 +245,7 @@ class SkyServiceSpec:
                         self.downscale_delay_seconds)
         add_if_not_none('load_balancing_policy', None,
                         self._load_balancing_policy)
+        add_if_not_none('ports', None, int(self.ports) if self.ports else None)
         if self.tls_credential is not None:
             add_if_not_none('tls', 'keyfile', self.tls_credential.keyfile)
             add_if_not_none('tls', 'certfile', self.tls_credential.certfile)
@@ -282,6 +293,9 @@ class SkyServiceSpec:
                 f'replica{max_plural} (target QPS per replica: '
                 f'{self.target_qps_per_replica})')
 
+    def set_ports(self, ports: str) -> None:
+        self._ports = ports
+
     def tls_str(self):
         if self.tls_credential is None:
             return 'No TLS Enabled'
@@ -320,6 +334,10 @@ class SkyServiceSpec:
         # If None, treated as having the same value of min_replicas.
         return self._max_replicas
 
+    @property
+    def ports(self) -> Optional[str]:
+        return self._ports
+
     @property
     def target_qps_per_replica(self) -> Optional[float]:
         return self._target_qps_per_replica

sky/utils/schemas.py

@@ -388,6 +388,9 @@ def get_service_schema():
                     },
                 }
             },
+            'ports': {
+                'type': 'integer',
+            },
             'replicas': {
                 'type': 'integer',
             },

{skypilot_nightly-1.0.0.dev20250129.dist-info → skypilot_nightly-1.0.0.dev20250130.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: skypilot-nightly
-Version: 1.0.0.dev20250129
+Version: 1.0.0.dev20250130
 Summary: SkyPilot: An intercloud broker for the clouds
 Author: SkyPilot Team
 License: Apache 2.0

{skypilot_nightly-1.0.0.dev20250129.dist-info → skypilot_nightly-1.0.0.dev20250130.dist-info}/RECORD

@@ -1,8 +1,8 @@
-sky/__init__.py,sha256=ExFZHFlVtSULSHNGhEbMwQWNbiaGjjVm4Q8JKdLDfWU,5529
+sky/__init__.py,sha256=_2mFMzFyJAHhW4BNdYQodCXMrGYYxEh_Dod5PwSOzJc,5529
 sky/admin_policy.py,sha256=hPo02f_A32gCqhUueF0QYy1fMSSKqRwYEg_9FxScN_s,3248
 sky/authentication.py,sha256=LXUDABKP1FJCS256xTTDJa40WXwHKF5x49S-4hZbD1M,21501
 sky/check.py,sha256=qTpm3N1zUZi2inEZPsrbt278B3h8nsk2gnepzIgLybE,10899
-sky/cli.py,sha256=xy5moD0W-retbQHxj7hq3q1mggU3dIWfepbj2Z8GCg0,213217
+sky/cli.py,sha256=-41DELQxAwIqLN3T1meUctCnjomujK9PK7njIiQUhKc,214883
 sky/cloud_stores.py,sha256=PcLT57_8SZy7o6paAluElfBynaLkbaOq3l-8dNg1AVM,23672
 sky/core.py,sha256=fE1rn4Ku94S0XmWTO5-6t6eT6aaJImNczRqEnTe8v7Q,38742
 sky/dag.py,sha256=f3sJlkH4bE6Uuz3ozNtsMhcBpRx7KmC9Sa4seDKt4hU,3104
@@ -99,7 +99,7 @@ sky/data/storage_utils.py,sha256=cM3kxlffYE7PnJySDu8huyUsMX_JYsf9uer8r5OYsjo,955
 sky/jobs/__init__.py,sha256=ObZcz3lL1ip8JcmR6gbfZ4RMMfXJJdsnuU2zLQUb8jY,1546
 sky/jobs/constants.py,sha256=6RphkJ6pmafQ7XYW5qwId1Zvqb99HJelA9kgrgfNR7o,1421
 sky/jobs/controller.py,sha256=0WcOk8xRZ-mZWuza-WE-ICKZTgZvXxNzj9pWXUslm6E,28312
-sky/jobs/core.py,sha256=2_Q9thiBPnd3i2nDqyUtQY-dsGZ1kRgAdnLcXHoycYo,19938
+sky/jobs/core.py,sha256=99fXKR-lhhFI4fTV7aGpbtbqgAkSgZBeUJ8Zt-68ar4,20314
 sky/jobs/recovery_strategy.py,sha256=m-EA-MWXPFrgx2CYFPr6MmgeUoDTEBmY2xruD2PRSGY,26365
 sky/jobs/scheduler.py,sha256=WAvNb8-vBk8q1zFordFdpH7gxqWDjPHDGZZay6aodOk,12028
 sky/jobs/state.py,sha256=bvBNZMg3DzPfS4eHNzMqYaMui2cqnWoWGDIaiOpaXSk,40770
@@ -191,14 +191,14 @@ sky/serve/__init__.py,sha256=Bqw8nB9u1QF3ryjbV797SPZq0DWAcjT94E_5B8J24ag,1808
 sky/serve/autoscalers.py,sha256=OxaynplCqbmrMA3fIGhxkugaGm-50QoI8S1fIfHK0M0,31667
 sky/serve/constants.py,sha256=7MflfgTHO9gDSux93U4BmNeEMWXxZB4q7I54KUwgp-s,4651
 sky/serve/controller.py,sha256=jtzWHsLHnVPQ727ZpDZTUpGTtIOssbnQpXeWOyAuW_s,11886
-sky/serve/core.py,sha256=RHPobivYdQj2PWs0uRLd_KvyAa7altqrT81si1IUlIs,35402
+sky/serve/core.py,sha256=rUCXibVGixO7j7b7nUcBY1pCcejOoa6tztIVf5MQ9bw,35778
 sky/serve/load_balancer.py,sha256=2nkMPRvy-h7hJL4Qq__tkT8nIAVC_nmjyXf8mMGYEFk,13658
 sky/serve/load_balancing_policies.py,sha256=XVj76qBgqh7h6wfx53RKQFzBefDWTE4TCdCEtFLLtI4,5398
-sky/serve/replica_managers.py,sha256=SFvK7ewilc3NVRcqXg63WtU1WmhJKPtJd27JfKR2aow,57716
+sky/serve/replica_managers.py,sha256=SW7k2iivUZ6dw_YMgGYOHOGD9_yyV4byfKa8e5t8_HE,57587
 sky/serve/serve_state.py,sha256=2V9WVD9OZ26vgyIJ-kZcrlouazwFTrpjlZAK9DRv3uA,20084
 sky/serve/serve_utils.py,sha256=m1Zcjslnzcr5AAppzV48WDOwMWjRaXotTUd_iN-dHgc,40654
 sky/serve/service.py,sha256=DPU1PJGuHa1WaNqxYqgpmqd4LA9jBbQM-KlLrA6C1M0,12156
-sky/serve/service_spec.py,sha256=26MCrv0mRbwfPCtg7uKD0xJjFXHzdajKM0gKuIGOKMc,16259
+sky/serve/service_spec.py,sha256=Q0qnFRjNnfGIpksubH5VqPKIlvpWs5had_Ma_PSHyo8,16940
 sky/setup_files/MANIFEST.in,sha256=WF0T89NLichHxZDDSQzvSpiONtAEFyur2MPmGczgTIo,555
 sky/setup_files/dependencies.py,sha256=OftFwWuV41sb_ZMD5euA6DABZx1Th1V_vCZcLV9CyMU,6234
 sky/setup_files/setup.py,sha256=HMqAIxHrhtQUOlm6_Iz5E_bL4dUvsYgXc9YVQIFayPs,7417
@@ -271,7 +271,7 @@ sky/utils/kubernetes_enums.py,sha256=imGqHSa8O07zD_6xH1SDMM7dBU5lF5fzFFlQuQy00QM
 sky/utils/log_utils.py,sha256=xEbUZfDiIiZkyWoLHXwIcqVMCBDEENsLCiogEXMDLt0,14139
 sky/utils/resources_utils.py,sha256=06Kx6AfbBdwBYGmIYFEY_qm6OBc2a5esZMPvIX7gCvc,7787
 sky/utils/rich_utils.py,sha256=hmnI1X5dKvRIQzB7EyNb34FT97qFNve-0QHqM5r0mVk,3066
-sky/utils/schemas.py,sha256=50bzZfjseNky8jUqmrjRYaOQQxkaM3mzQYyLiRq0qfs,30482
+sky/utils/schemas.py,sha256=32w70vJN0-C-00--QPGw_8zQTRnpgDMH-eSuxPII4bM,30555
 sky/utils/subprocess_utils.py,sha256=YhtxqHoaZLw2M9TikTH56dTboZN3Qu2RsGeWo4uwJVA,12054
 sky/utils/timeline.py,sha256=ebHxKJK2HX0utGArrUgSezTPkcwav3VETa_AQS34t-E,3925
 sky/utils/ux_utils.py,sha256=CqyIFGDuSE8fQasPkna_loZMwtboC9KedR09WEQ7qz0,6502
@@ -289,9 +289,9 @@ sky/utils/kubernetes/k8s_gpu_labeler_job.yaml,sha256=k0TBoQ4zgf79-sVkixKSGYFHQ7Z
 sky/utils/kubernetes/k8s_gpu_labeler_setup.yaml,sha256=VLKT2KKimZu1GDg_4AIlIt488oMQvhRZWwsj9vBbPUg,3812
 sky/utils/kubernetes/rsync_helper.sh,sha256=h4YwrPFf9727CACnMJvF3EyK_0OeOYKKt4su_daKekw,1256
 sky/utils/kubernetes/ssh_jump_lifecycle_manager.py,sha256=Kq1MDygF2IxFmu9FXpCxqucXLmeUrvs6OtRij6XTQbo,6554
-skypilot_nightly-1.0.0.dev20250129.dist-info/LICENSE,sha256=emRJAvE7ngL6x0RhQvlns5wJzGI3NEQ_WMjNmd9TZc4,12170
-skypilot_nightly-1.0.0.dev20250129.dist-info/METADATA,sha256=t06vBK0dA7BIxwSsBsD477mo_-DgGUXnt4etMTMJQDw,21249
-skypilot_nightly-1.0.0.dev20250129.dist-info/WHEEL,sha256=In9FTNxeP60KnTkGw7wk6mJPYd_dQSjEZmXdBdMCI-8,91
-skypilot_nightly-1.0.0.dev20250129.dist-info/entry_points.txt,sha256=StA6HYpuHj-Y61L2Ze-hK2IcLWgLZcML5gJu8cs6nU4,36
-skypilot_nightly-1.0.0.dev20250129.dist-info/top_level.txt,sha256=qA8QuiNNb6Y1OF-pCUtPEr6sLEwy2xJX06Bd_CrtrHY,4
-skypilot_nightly-1.0.0.dev20250129.dist-info/RECORD,,
+skypilot_nightly-1.0.0.dev20250130.dist-info/LICENSE,sha256=emRJAvE7ngL6x0RhQvlns5wJzGI3NEQ_WMjNmd9TZc4,12170
+skypilot_nightly-1.0.0.dev20250130.dist-info/METADATA,sha256=h7cYgI0Xy2rdgSRNc8aciOdrnEJ4oqZuAlnCy0N9xb8,21249
+skypilot_nightly-1.0.0.dev20250130.dist-info/WHEEL,sha256=In9FTNxeP60KnTkGw7wk6mJPYd_dQSjEZmXdBdMCI-8,91
+skypilot_nightly-1.0.0.dev20250130.dist-info/entry_points.txt,sha256=StA6HYpuHj-Y61L2Ze-hK2IcLWgLZcML5gJu8cs6nU4,36
+skypilot_nightly-1.0.0.dev20250130.dist-info/top_level.txt,sha256=qA8QuiNNb6Y1OF-pCUtPEr6sLEwy2xJX06Bd_CrtrHY,4
+skypilot_nightly-1.0.0.dev20250130.dist-info/RECORD,,