skypilot-nightly 1.0.0.dev20250926__tar.gz → 1.0.0.dev20251001__tar.gz

{skypilot_nightly-1.0.0.dev20250926/skypilot_nightly.egg-info → skypilot_nightly-1.0.0.dev20251001}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: skypilot-nightly
-Version: 1.0.0.dev20250926
+Version: 1.0.0.dev20251001
 Summary: SkyPilot: Run AI on Any Infra — Unified, Faster, Cheaper.
 Author: SkyPilot Team
 License: Apache 2.0
@@ -154,51 +154,51 @@ Requires-Dist: protobuf<7.0.0,>=5.26.1; extra == "server"
 Requires-Dist: aiosqlite; extra == "server"
 Requires-Dist: greenlet; extra == "server"
 Provides-Extra: all
-Requires-Dist: botocore>=1.29.10; extra == "all"
-Requires-Dist: pyjwt; extra == "all"
+Requires-Dist: anyio; extra == "all"
+Requires-Dist: grpcio>=1.63.0; extra == "all"
+Requires-Dist: protobuf<7.0.0,>=5.26.1; extra == "all"
+Requires-Dist: sqlalchemy_adapter; extra == "all"
+Requires-Dist: cudo-compute>=0.1.10; extra == "all"
+Requires-Dist: ibm-cloud-sdk-core; extra == "all"
+Requires-Dist: casbin; extra == "all"
+Requires-Dist: awscli>=1.27.10; extra == "all"
+Requires-Dist: azure-common; extra == "all"
 Requires-Dist: azure-storage-blob>=12.23.1; extra == "all"
 Requires-Dist: vastai-sdk>=0.1.12; extra == "all"
-Requires-Dist: azure-identity>=1.19.0; extra == "all"
-Requires-Dist: azure-mgmt-compute>=33.0.0; extra == "all"
-Requires-Dist: ibm-cloud-sdk-core; extra == "all"
-Requires-Dist: google-cloud-storage; extra == "all"
-Requires-Dist: passlib; extra == "all"
+Requires-Dist: aiohttp; extra == "all"
 Requires-Dist: greenlet; extra == "all"
-Requires-Dist: azure-cli>=2.65.0; extra == "all"
-Requires-Dist: oci; extra == "all"
-Requires-Dist: cudo-compute>=0.1.10; extra == "all"
+Requires-Dist: pyjwt; extra == "all"
+Requires-Dist: nebius>=0.2.47; extra == "all"
+Requires-Dist: azure-core>=1.24.0; extra == "all"
+Requires-Dist: runpod>=1.6.1; extra == "all"
+Requires-Dist: msgraph-sdk; extra == "all"
+Requires-Dist: botocore>=1.29.10; extra == "all"
+Requires-Dist: azure-identity>=1.19.0; extra == "all"
+Requires-Dist: pydo>=0.3.0; extra == "all"
 Requires-Dist: websockets; extra == "all"
-Requires-Dist: azure-core>=1.31.0; extra == "all"
-Requires-Dist: ray[default]>=2.6.1; extra == "all"
+Requires-Dist: pyopenssl<24.3.0,>=23.2.0; extra == "all"
+Requires-Dist: tomli; python_version < "3.11" and extra == "all"
+Requires-Dist: azure-mgmt-compute>=33.0.0; extra == "all"
 Requires-Dist: boto3>=1.26.1; extra == "all"
-Requires-Dist: ibm-vpc; extra == "all"
+Requires-Dist: aiosqlite; extra == "all"
+Requires-Dist: azure-core>=1.31.0; extra == "all"
+Requires-Dist: colorama<0.4.5; extra == "all"
+Requires-Dist: docker; extra == "all"
+Requires-Dist: azure-mgmt-network>=27.0.0; extra == "all"
 Requires-Dist: pyvmomi==8.0.1.0.2; extra == "all"
-Requires-Dist: azure-common; extra == "all"
+Requires-Dist: kubernetes!=32.0.0,>=20.0.0; extra == "all"
+Requires-Dist: ibm-cos-sdk; extra == "all"
 Requires-Dist: google-api-python-client>=2.69.0; extra == "all"
+Requires-Dist: ibm-vpc; extra == "all"
 Requires-Dist: msrestazure; extra == "all"
-Requires-Dist: ibm-platform-services>=0.48.0; extra == "all"
-Requires-Dist: tomli; python_version < "3.11" and extra == "all"
-Requires-Dist: msgraph-sdk; extra == "all"
-Requires-Dist: ibm-cos-sdk; extra == "all"
-Requires-Dist: awscli>=1.27.10; extra == "all"
-Requires-Dist: azure-core>=1.24.0; extra == "all"
-Requires-Dist: anyio; extra == "all"
-Requires-Dist: pyopenssl<24.3.0,>=23.2.0; extra == "all"
-Requires-Dist: aiosqlite; extra == "all"
+Requires-Dist: passlib; extra == "all"
+Requires-Dist: oci; extra == "all"
 Requires-Dist: ecsapi>=0.2.0; extra == "all"
-Requires-Dist: protobuf<7.0.0,>=5.26.1; extra == "all"
-Requires-Dist: grpcio>=1.63.0; extra == "all"
-Requires-Dist: pydo>=0.3.0; extra == "all"
-Requires-Dist: casbin; extra == "all"
-Requires-Dist: aiohttp; extra == "all"
-Requires-Dist: kubernetes!=32.0.0,>=20.0.0; extra == "all"
+Requires-Dist: ray[default]>=2.6.1; extra == "all"
+Requires-Dist: ibm-platform-services>=0.48.0; extra == "all"
+Requires-Dist: azure-cli>=2.65.0; extra == "all"
+Requires-Dist: google-cloud-storage; extra == "all"
 Requires-Dist: python-dateutil; extra == "all"
-Requires-Dist: nebius>=0.2.47; extra == "all"
-Requires-Dist: docker; extra == "all"
-Requires-Dist: colorama<0.4.5; extra == "all"
-Requires-Dist: runpod>=1.6.1; extra == "all"
-Requires-Dist: azure-mgmt-network>=27.0.0; extra == "all"
-Requires-Dist: sqlalchemy_adapter; extra == "all"
 Dynamic: author
 Dynamic: classifier
 Dynamic: description

{skypilot_nightly-1.0.0.dev20250926 → skypilot_nightly-1.0.0.dev20251001}/sky/__init__.py

@@ -7,7 +7,7 @@ import urllib.request
 from sky.utils import directory_utils
 
 # Replaced with the current commit when building the wheels.
-_SKYPILOT_COMMIT_SHA = '827d534c8bbfa61b895467b9431283e923dd9841'
+_SKYPILOT_COMMIT_SHA = '047e366ad4c073a5753b62cc5d4c3c4660b1476a'
 
 
 def _get_git_commit():
@@ -37,7 +37,7 @@ def _get_git_commit():
 
 
 __commit__ = _get_git_commit()
-__version__ = '1.0.0.dev20250926'
+__version__ = '1.0.0.dev20251001'
 __root_dir__ = directory_utils.get_sky_dir()
 
 

{skypilot_nightly-1.0.0.dev20250926 → skypilot_nightly-1.0.0.dev20251001}/sky/backends/backend_utils.py

@@ -16,8 +16,8 @@ import tempfile
 import threading
 import time
 import typing
-from typing import (Any, Callable, Dict, List, Optional, Sequence, Set, Tuple,
-                    TypeVar, Union)
+from typing import (Any, Callable, Dict, Iterator, List, Optional, Sequence,
+                    Set, Tuple, TypeVar, Union)
 import uuid
 
 import aiohttp
@@ -723,11 +723,15 @@ def write_cluster_config(
                 'is not supported by this cloud. Remove the config or set: '
                 '`remote_identity: LOCAL_CREDENTIALS`.')
         if isinstance(cloud, clouds.Kubernetes):
-            if skypilot_config.get_effective_region_config(
+            allowed_contexts = skypilot_config.get_workspace_cloud(
+                'kubernetes').get('allowed_contexts', None)
+            if allowed_contexts is None:
+                allowed_contexts = skypilot_config.get_effective_region_config(
                     cloud='kubernetes',
                     region=None,
                     keys=('allowed_contexts',),
-                    default_value=None) is None:
+                    default_value=None)
+            if allowed_contexts is None:
                 excluded_clouds.add(cloud)
         else:
             excluded_clouds.add(cloud)
@@ -1226,7 +1230,6 @@ def _deterministic_cluster_yaml_hash(tmp_yaml_path: str) -> str:
     Rather than constructing the whole byte sequence, which may be quite large,
     we construct it incrementally by using hash.update() to add new bytes.
     """
-
     # Load the yaml contents so that we can directly remove keys.
     yaml_config = yaml_utils.read_yaml(tmp_yaml_path)
     for key_list in _RAY_YAML_KEYS_TO_REMOVE_FOR_HASH:
@@ -2614,7 +2617,7 @@ def refresh_cluster_record(
         cluster_name: str,
         *,
         force_refresh_statuses: Optional[Set[status_lib.ClusterStatus]] = None,
-        acquire_per_cluster_status_lock: bool = True,
+        cluster_lock_already_held: bool = False,
         cluster_status_lock_timeout: int = CLUSTER_STATUS_LOCK_TIMEOUT_SECONDS,
         include_user_info: bool = True,
         summary_response: bool = False) -> Optional[Dict[str, Any]]:
@@ -2634,9 +2637,13 @@ def refresh_cluster_record(
               _CLUSTER_STATUS_CACHE_DURATION_SECONDS old, and one of:
                 1. the cluster is a spot cluster, or
                 2. cluster autostop is set and the cluster is not STOPPED.
-        acquire_per_cluster_status_lock: Whether to acquire the per-cluster lock
-          before updating the status. Even if this is True, the lock may not be
-          acquired if the status does not need to be refreshed.
+        cluster_lock_already_held: Whether the caller is already holding the
+          per-cluster lock. You MUST NOT set this to True if the caller does not
+          already hold the lock. If True, we will not acquire the lock before
+          updating the status. Failing to hold the lock while updating the
+          status can lead to correctness issues - e.g. an launch in-progress may
+          appear to be DOWN incorrectly. Even if this is set to False, the lock
+          may not be acquired if the status does not need to be refreshed.
         cluster_status_lock_timeout: The timeout to acquire the per-cluster
           lock. If timeout, the function will use the cached status. If the
           value is <0, do not timeout (wait for the lock indefinitely). By
@@ -2687,7 +2694,7 @@ def refresh_cluster_record(
             if not _must_refresh_cluster_status(record, force_refresh_statuses):
                 return record
 
-            if not acquire_per_cluster_status_lock:
+            if cluster_lock_already_held:
                 return _update_cluster_status(cluster_name, include_user_info,
                                               summary_response)
 
@@ -2741,7 +2748,7 @@ def refresh_cluster_status_handle(
     cluster_name: str,
     *,
     force_refresh_statuses: Optional[Set[status_lib.ClusterStatus]] = None,
-    acquire_per_cluster_status_lock: bool = True,
+    cluster_lock_already_held: bool = False,
     cluster_status_lock_timeout: int = CLUSTER_STATUS_LOCK_TIMEOUT_SECONDS
 ) -> Tuple[Optional[status_lib.ClusterStatus],
            Optional[backends.ResourceHandle]]:
@@ -2754,7 +2761,7 @@ def refresh_cluster_status_handle(
     record = refresh_cluster_record(
         cluster_name,
         force_refresh_statuses=force_refresh_statuses,
-        acquire_per_cluster_status_lock=acquire_per_cluster_status_lock,
+        cluster_lock_already_held=cluster_lock_already_held,
         cluster_status_lock_timeout=cluster_status_lock_timeout,
         include_user_info=False,
         summary_response=True)
@@ -3079,7 +3086,7 @@ def _refresh_cluster(
         record = refresh_cluster_record(
             cluster_name,
             force_refresh_statuses=force_refresh_statuses,
-            acquire_per_cluster_status_lock=True,
+            cluster_lock_already_held=False,
             include_user_info=include_user_info,
             summary_response=summary_response)
     except (exceptions.ClusterStatusFetchingError,
@@ -3859,13 +3866,35 @@ def invoke_skylet_with_retries(func: Callable[..., T]) -> T:
     ) from last_exception
 
 
+def invoke_skylet_streaming_with_retries(
+        stream_func: Callable[..., Iterator[T]]) -> Iterator[T]:
+    """Generic helper for making Skylet streaming gRPC requests."""
+    max_attempts = 3
+    backoff = common_utils.Backoff(initial_backoff=0.5)
+    last_exception: Optional[Exception] = None
+
+    for _ in range(max_attempts):
+        try:
+            for response in stream_func():
+                yield response
+            return
+        except grpc.RpcError as e:
+            last_exception = e
+            _handle_grpc_error(e, backoff.current_backoff())
+
+    raise RuntimeError(
+        f'Failed to stream Skylet response after {max_attempts} attempts'
+    ) from last_exception
+
+
 def _handle_grpc_error(e: 'grpc.RpcError', current_backoff: float) -> None:
     if e.code() == grpc.StatusCode.INTERNAL:
         with ux_utils.print_exception_no_traceback():
             raise exceptions.SkyletInternalError(e.details())
     elif e.code() == grpc.StatusCode.UNAVAILABLE:
         time.sleep(current_backoff)
-    elif e.code() == grpc.StatusCode.UNIMPLEMENTED:
+    elif e.code() == grpc.StatusCode.UNIMPLEMENTED or e.code(
+    ) == grpc.StatusCode.UNKNOWN:
         # Handle backwards compatibility: old server doesn't implement this RPC.
         # Let the caller fall back to legacy execution.
         raise exceptions.SkyletMethodNotImplementedError(

{skypilot_nightly-1.0.0.dev20250926 → skypilot_nightly-1.0.0.dev20251001}/sky/backends/cloud_vm_ray_backend.py

@@ -19,8 +19,8 @@ import textwrap
 import threading
 import time
 import typing
-from typing import (Any, Callable, Dict, Iterable, List, Optional, Set, Tuple,
-                    Union)
+from typing import (Any, Callable, Dict, Iterable, Iterator, List, Optional,
+                    Set, Tuple, Union)
 
 import colorama
 import psutil
@@ -91,6 +91,8 @@ if typing.TYPE_CHECKING:
     from sky.schemas.generated import autostopv1_pb2_grpc
     from sky.schemas.generated import jobsv1_pb2
     from sky.schemas.generated import jobsv1_pb2_grpc
+    from sky.schemas.generated import managed_jobsv1_pb2
+    from sky.schemas.generated import managed_jobsv1_pb2_grpc
     from sky.schemas.generated import servev1_pb2
     from sky.schemas.generated import servev1_pb2_grpc
 else:
@@ -111,6 +113,10 @@ else:
         'sky.schemas.generated.servev1_pb2')
     servev1_pb2_grpc = adaptors_common.LazyImport(
         'sky.schemas.generated.servev1_pb2_grpc')
+    managed_jobsv1_pb2 = adaptors_common.LazyImport(
+        'sky.schemas.generated.managed_jobsv1_pb2')
+    managed_jobsv1_pb2_grpc = adaptors_common.LazyImport(
+        'sky.schemas.generated.managed_jobsv1_pb2_grpc')
 
 Path = str
 
@@ -2737,6 +2743,11 @@ class CloudVmRayResourceHandle(backends.backend.ResourceHandle):
             (tunnel.port, tunnel.pid) if tunnel is not None else None)
 
     def get_grpc_channel(self) -> 'grpc.Channel':
+        grpc_options = [
+            # The task YAMLs can be large, so the default
+            # max_receive_message_length of 4MB might not be enough.
+            ('grpc.max_receive_message_length', -1),
+        ]
         # It's fine to not grab the lock here, as we're only reading,
         # and writes are very rare.
         # It's acceptable to read while another process is opening a tunnel,
@@ -2753,7 +2764,8 @@ class CloudVmRayResourceHandle(backends.backend.ResourceHandle):
                 with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
                     s.settimeout(0.5)
                     s.connect(('localhost', tunnel.port))
-                return grpc.insecure_channel(f'localhost:{tunnel.port}')
+                return grpc.insecure_channel(f'localhost:{tunnel.port}',
+                                             options=grpc_options)
             except socket.error as e:
                 logger.warning(
                     'Failed to connect to SSH tunnel for cluster '
@@ -2772,19 +2784,22 @@ class CloudVmRayResourceHandle(backends.backend.ResourceHandle):
                                  f'{self.cluster_name!r}, '
                                  'opening the tunnel')
                     tunnel = self._open_and_update_skylet_tunnel()
-                    return grpc.insecure_channel(f'localhost:{tunnel.port}')
+                    return grpc.insecure_channel(f'localhost:{tunnel.port}',
+                                                 options=grpc_options)
                 try:
                     with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
                         s.settimeout(0.5)
                         s.connect(('localhost', tunnel.port))
-                        return grpc.insecure_channel(f'localhost:{tunnel.port}')
+                        return grpc.insecure_channel(f'localhost:{tunnel.port}',
+                                                     options=grpc_options)
                 except socket.error as e:
                     logger.warning(
                         'Failed to connect to SSH tunnel for cluster '
                         f'{self.cluster_name!r} on port {tunnel.port} ({e}), '
                         'opening new tunnel')
                     tunnel = self._open_and_update_skylet_tunnel()
-                    return grpc.insecure_channel(f'localhost:{tunnel.port}')
+                    return grpc.insecure_channel(f'localhost:{tunnel.port}',
+                                                 options=grpc_options)
         except locks.LockTimeout as e:
             raise RuntimeError(
                 'Failed to get gRPC channel for cluster '
@@ -3060,6 +3075,8 @@ class SkyletClient:
         self._autostop_stub = autostopv1_pb2_grpc.AutostopServiceStub(channel)
         self._jobs_stub = jobsv1_pb2_grpc.JobsServiceStub(channel)
         self._serve_stub = servev1_pb2_grpc.ServeServiceStub(channel)
+        self._managed_jobs_stub = (
+            managed_jobsv1_pb2_grpc.ManagedJobsServiceStub(channel))
 
     def set_autostop(
         self,
@@ -3146,6 +3163,13 @@ class SkyletClient:
     ) -> 'jobsv1_pb2.GetLogDirsForJobsResponse':
         return self._jobs_stub.GetLogDirsForJobs(request, timeout=timeout)
 
+    def tail_logs(
+        self,
+        request: 'jobsv1_pb2.TailLogsRequest',
+        timeout: Optional[float] = constants.SKYLET_GRPC_TIMEOUT_SECONDS
+    ) -> Iterator['jobsv1_pb2.TailLogsResponse']:
+        return self._jobs_stub.TailLogs(request, timeout=timeout)
+
     def get_service_status(
         self,
         request: 'servev1_pb2.GetServiceStatusRequest',
@@ -3194,6 +3218,35 @@ class SkyletClient:
     ) -> 'servev1_pb2.UpdateServiceResponse':
         return self._serve_stub.UpdateService(request, timeout=timeout)
 
+    def get_managed_job_controller_version(
+        self,
+        request: 'managed_jobsv1_pb2.GetVersionRequest',
+        timeout: Optional[float] = constants.SKYLET_GRPC_TIMEOUT_SECONDS
+    ) -> 'managed_jobsv1_pb2.GetVersionResponse':
+        return self._managed_jobs_stub.GetVersion(request, timeout=timeout)
+
+    def get_managed_job_table(
+        self,
+        request: 'managed_jobsv1_pb2.GetJobTableRequest',
+        timeout: Optional[float] = constants.SKYLET_GRPC_TIMEOUT_SECONDS
+    ) -> 'managed_jobsv1_pb2.GetJobTableResponse':
+        return self._managed_jobs_stub.GetJobTable(request, timeout=timeout)
+
+    def get_all_managed_job_ids_by_name(
+        self,
+        request: 'managed_jobsv1_pb2.GetAllJobIdsByNameRequest',
+        timeout: Optional[float] = constants.SKYLET_GRPC_TIMEOUT_SECONDS
+    ) -> 'managed_jobsv1_pb2.GetAllJobIdsByNameResponse':
+        return self._managed_jobs_stub.GetAllJobIdsByName(request,
+                                                          timeout=timeout)
+
+    def cancel_managed_jobs(
+        self,
+        request: 'managed_jobsv1_pb2.CancelJobsRequest',
+        timeout: Optional[float] = constants.SKYLET_GRPC_TIMEOUT_SECONDS
+    ) -> 'managed_jobsv1_pb2.CancelJobsResponse':
+        return self._managed_jobs_stub.CancelJobs(request, timeout=timeout)
+
 
 @registry.BACKEND_REGISTRY.type_register(name='cloudvmray')
 class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
@@ -3706,7 +3759,7 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
                     ux_utils.spinner_message('Preparing SkyPilot runtime')):
                 use_legacy = not handle.is_grpc_enabled_with_flag
 
-                if handle.is_grpc_enabled_with_flag:
+                if not use_legacy:
                     try:
                         request = jobsv1_pb2.UpdateStatusRequest()
                         backend_utils.invoke_skylet_with_retries(
@@ -3730,7 +3783,7 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
             # 2. On next `sky start`, it gets reset to FAILED.
             use_legacy = not handle.is_grpc_enabled_with_flag
 
-            if handle.is_grpc_enabled_with_flag:
+            if not use_legacy:
                 try:
                     fail_request = jobsv1_pb2.FailAllInProgressJobsRequest()
                     backend_utils.invoke_skylet_with_retries(
@@ -4165,7 +4218,7 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
             _dump_code_to_file(job_submit_cmd,
                                constants.PERSISTENT_RUN_SCRIPT_DIR)
 
-        if handle.is_grpc_enabled_with_flag:
+        if not use_legacy:
             try:
                 managed_job_info: Optional[jobsv1_pb2.ManagedJobInfo] = None
                 if managed_job_dag is not None:
@@ -4297,7 +4350,7 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
                  metadata: str) -> Tuple[int, str]:
         use_legacy = not handle.is_grpc_enabled_with_flag
 
-        if handle.is_grpc_enabled_with_flag:
+        if not use_legacy:
             try:
                 request = jobsv1_pb2.AddJobRequest(
                     job_name=job_name,
@@ -4567,7 +4620,7 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
         """
         use_legacy = not handle.is_grpc_enabled_with_flag
 
-        if handle.is_grpc_enabled_with_flag:
+        if not use_legacy:
             try:
                 request = jobsv1_pb2.CancelJobsRequest(job_ids=jobs,
                                                        cancel_all=cancel_all,
@@ -4610,7 +4663,7 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
         job_to_dir: Dict[str, str] = {}
         use_legacy = not handle.is_grpc_enabled_with_flag
 
-        if handle.is_grpc_enabled_with_flag:
+        if not use_legacy:
             try:
                 int_job_ids = []
                 if job_ids:
@@ -4724,6 +4777,28 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
             The exit code of the tail command. Returns code 100 if the job has
             failed. See exceptions.JobExitCode for possible return codes.
         """
+        if handle.is_grpc_enabled_with_flag:
+            last_exit_code = 0
+            try:
+                request = jobsv1_pb2.TailLogsRequest(
+                    job_id=job_id,
+                    managed_job_id=managed_job_id,
+                    follow=follow,
+                    tail=tail)
+                for resp in backend_utils.invoke_skylet_streaming_with_retries(
+                        lambda: SkyletClient(handle.get_grpc_channel()
+                                            ).tail_logs(request, timeout=None)):
+                    if resp.log_line:
+                        print(resp.log_line, end='', flush=True)
+                    last_exit_code = resp.exit_code
+                return last_exit_code
+            except exceptions.SkyletMethodNotImplementedError:
+                pass
+            except grpc.RpcError as e:
+                if e.code() == grpc.StatusCode.CANCELLED:
+                    return last_exit_code
+                raise e
+
         code = job_lib.JobLibCodeGen.tail_logs(job_id,
                                                managed_job_id=managed_job_id,
                                                follow=follow,
@@ -4761,6 +4836,7 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
                               tail: Optional[int] = None) -> int:
         # if job_name is not None, job_id should be None
         assert job_name is None or job_id is None, (job_name, job_id)
+        # TODO(kevin): Migrate stream_logs to gRPC
         code = managed_jobs.ManagedJobCodeGen.stream_logs(
             job_name, job_id, follow, controller, tail)
 
@@ -4806,20 +4882,37 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
         assert job_name is None or job_id is None, (job_name, job_id)
 
         if job_id is None:
-            # generate code to get the job_id
+            # get the job_id
             # if job_name is None, get all job_ids
             # TODO: Only get the latest job_id, since that's the only one we use
-            code = managed_jobs.ManagedJobCodeGen.get_all_job_ids_by_name(
-                job_name=job_name)
-            returncode, job_ids, stderr = self.run_on_head(handle,
-                                                           code,
-                                                           stream_logs=False,
-                                                           require_outputs=True,
-                                                           separate_stderr=True)
-            subprocess_utils.handle_returncode(returncode, code,
-                                               'Failed to sync down logs.',
-                                               stderr)
-            job_ids = message_utils.decode_payload(job_ids)
+
+            use_legacy = not handle.is_grpc_enabled_with_flag
+            logger.info(f'handle.is_grpc_enabled_with_flag: '
+                        f'{handle.is_grpc_enabled_with_flag}')
+            if not use_legacy:
+                try:
+                    request = managed_jobsv1_pb2.GetAllJobIdsByNameRequest(
+                        job_name=job_name)
+                    response = backend_utils.invoke_skylet_with_retries(
+                        lambda: SkyletClient(handle.get_grpc_channel(
+                        )).get_all_managed_job_ids_by_name(request))
+                    job_ids = list(response.job_ids)
+                except exceptions.SkyletMethodNotImplementedError:
+                    use_legacy = True
+
+            if use_legacy:
+                code = managed_jobs.ManagedJobCodeGen.get_all_job_ids_by_name(
+                    job_name=job_name)
+                returncode, job_ids_payload, stderr = self.run_on_head(
+                    handle,
+                    code,
+                    stream_logs=False,
+                    require_outputs=True,
+                    separate_stderr=True)
+                subprocess_utils.handle_returncode(returncode, code,
+                                                   'Failed to sync down logs.',
+                                                   stderr)
+                job_ids = message_utils.decode_payload(job_ids_payload)
             if not job_ids:
                 logger.info(f'{colorama.Fore.YELLOW}'
                             'No matching job found'
@@ -4847,18 +4940,39 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
         else:
             # get the run_timestamp
             # the function takes in [job_id]
-            code = job_lib.JobLibCodeGen.get_log_dirs_for_jobs([str(job_id)])
-            returncode, run_timestamps_payload, stderr = self.run_on_head(
-                handle,
-                code,
-                stream_logs=False,
-                require_outputs=True,
-                separate_stderr=True)
-            subprocess_utils.handle_returncode(returncode, code,
-                                               'Failed to sync logs.', stderr)
-            # returns with a dict of {job_id: run_timestamp}
-            run_timestamps = message_utils.decode_payload(
-                run_timestamps_payload)
+            use_legacy = not handle.is_grpc_enabled_with_flag
+            if not use_legacy:
+                try:
+                    log_dirs_request = jobsv1_pb2.GetLogDirsForJobsRequest(
+                        job_ids=[job_id])
+                    log_dirs_response = (
+                        backend_utils.invoke_skylet_with_retries(
+                            lambda: SkyletClient(handle.get_grpc_channel(
+                            )).get_log_dirs_for_jobs(log_dirs_request)))
+                    job_log_dirs = log_dirs_response.job_log_dirs
+                    # Convert back to the expected format
+                    # {job_id: run_timestamp}
+                    run_timestamps = {}
+                    for jid, log_dir in job_log_dirs.items():
+                        run_timestamps[int(jid)] = log_dir
+                except exceptions.SkyletMethodNotImplementedError:
+                    use_legacy = True
+
+            if use_legacy:
+                code = job_lib.JobLibCodeGen.get_log_dirs_for_jobs(
+                    [str(job_id)])
+                returncode, run_timestamps_payload, stderr = self.run_on_head(
+                    handle,
+                    code,
+                    stream_logs=False,
+                    require_outputs=True,
+                    separate_stderr=True)
+                subprocess_utils.handle_returncode(returncode, code,
+                                                   'Failed to sync logs.',
+                                                   stderr)
+                # returns with a dict of {job_id: run_timestamp}
+                run_timestamps = message_utils.decode_payload(
+                    run_timestamps_payload)
         if not run_timestamps:
             logger.info(f'{colorama.Fore.YELLOW}'
                         'No matching log directories found'
@@ -4925,6 +5039,7 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
                         exist_ok=True)
             log_file = os.path.join(local_log_dir, 'run.log')
 
+            # TODO(kevin): Migrate stream_logs to gRPC
             code = managed_jobs.ManagedJobCodeGen.stream_logs(
                 job_name=None,
                 job_id=int(job_id),
@@ -5006,7 +5121,7 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
                         # observed in AWS. See also
                         # _LAUNCH_DOUBLE_CHECK_WINDOW in backend_utils.py.
                         force_refresh_statuses={status_lib.ClusterStatus.INIT},
-                        acquire_per_cluster_status_lock=False))
+                        cluster_lock_already_held=True))
                 cluster_status_fetched = True
             except exceptions.ClusterStatusFetchingError:
                 logger.warning(
@@ -5610,7 +5725,7 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
             record = backend_utils.refresh_cluster_record(
                 cluster_name,
                 force_refresh_statuses={status_lib.ClusterStatus.INIT},
-                acquire_per_cluster_status_lock=False,
+                cluster_lock_already_held=True,
                 include_user_info=False,
                 summary_response=True,
             )

{skypilot_nightly-1.0.0.dev20250926 → skypilot_nightly-1.0.0.dev20251001}/sky/check.py

@@ -621,35 +621,6 @@ def _format_enabled_cloud(cloud_name: str,
     if cloud_name in [repr(sky_clouds.Kubernetes()), repr(sky_clouds.SSH())]:
         return (f'{title}' + _format_context_details(
             cloud_name, show_details=False, ctx2text=ctx2text))
-
-    if cloud_name == repr(sky_clouds.Kubernetes()):
-        # Get enabled contexts for Kubernetes
-        existing_contexts = sky_clouds.Kubernetes.existing_allowed_contexts()
-        if not existing_contexts:
-            return _green_color(cloud_and_capabilities)
-
-        # Check if allowed_contexts is explicitly set in config
-        allowed_contexts = skypilot_config.get_effective_region_config(
-            cloud='kubernetes',
-            region=None,
-            keys=('allowed_contexts',),
-            default_value=None)
-
-        # Format the context info with consistent styling
-        if allowed_contexts is not None:
-            contexts_formatted = []
-            for i, context in enumerate(existing_contexts):
-                symbol = (ux_utils.INDENT_LAST_SYMBOL
-                          if i == len(existing_contexts) -
-                          1 else ux_utils.INDENT_SYMBOL)
-                contexts_formatted.append(f'\n    {symbol}{context}')
-            context_info = f'  Allowed contexts:{"".join(contexts_formatted)}'
-        else:
-            context_info = f'  Active context: {existing_contexts[0]}'
-
-        return (f'{_green_color(cloud_and_capabilities)}\n'
-                f'  {colorama.Style.DIM}{context_info}'
-                f'{colorama.Style.RESET_ALL}')
     return _green_color(cloud_and_capabilities)