skypilot-nightly 1.0.0.dev20250320__tar.gz → 1.0.0.dev20250322__tar.gz

{skypilot_nightly-1.0.0.dev20250320/skypilot_nightly.egg-info → skypilot_nightly-1.0.0.dev20250322}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: skypilot-nightly
-Version: 1.0.0.dev20250320
+Version: 1.0.0.dev20250322
 Summary: SkyPilot: An intercloud broker for the clouds
 Author: SkyPilot Team
 License: Apache 2.0
@@ -108,6 +108,11 @@ Provides-Extra: vsphere
 Requires-Dist: pyvmomi==8.0.1.0.2; extra == "vsphere"
 Provides-Extra: nebius
 Requires-Dist: nebius>=0.2.0; extra == "nebius"
+Requires-Dist: urllib3<2; extra == "nebius"
+Requires-Dist: awscli>=1.27.10; extra == "nebius"
+Requires-Dist: botocore>=1.29.10; extra == "nebius"
+Requires-Dist: boto3>=1.26.1; extra == "nebius"
+Requires-Dist: colorama<0.4.5; extra == "nebius"
 Provides-Extra: all
 Requires-Dist: urllib3<2; extra == "all"
 Requires-Dist: awscli>=1.27.10; extra == "all"
@@ -151,6 +156,11 @@ Requires-Dist: azure-common; extra == "all"
 Requires-Dist: vastai-sdk>=0.1.12; extra == "all"
 Requires-Dist: pyvmomi==8.0.1.0.2; extra == "all"
 Requires-Dist: nebius>=0.2.0; extra == "all"
+Requires-Dist: urllib3<2; extra == "all"
+Requires-Dist: awscli>=1.27.10; extra == "all"
+Requires-Dist: botocore>=1.29.10; extra == "all"
+Requires-Dist: boto3>=1.26.1; extra == "all"
+Requires-Dist: colorama<0.4.5; extra == "all"
 Dynamic: author
 Dynamic: classifier
 Dynamic: description
@@ -222,8 +232,8 @@ SkyPilot **unifies multiple clusters, clouds, and hardware**:
 
 SkyPilot **cuts your cloud costs & maximizes GPU availability**:
 * Autostop: automatic cleanup of idle resources
-* [Managed Spot](https://docs.skypilot.co/en/latest/examples/managed-jobs.html): 3-6x cost savings using spot instances, with preemption auto-recovery
-* Optimizer: auto-selects the cheapest & most available infra
+* [Spot instance support](https://docs.skypilot.co/en/latest/examples/managed-jobs.html#running-on-spot-instances): 3-6x cost savings, with preemption auto-recovery
+* Intelligent scheduling: automatically run on the cheapest & most available infra
 
 SkyPilot supports your existing GPU, TPU, and CPU workloads, with no code changes.
 
@@ -238,6 +248,9 @@ To get the latest features and fixes, use the nightly build or [install from sou
 pip install "skypilot-nightly[kubernetes,aws,gcp,azure,oci,lambda,runpod,fluidstack,paperspace,cudo,ibm,scp,nebius]"
 ```
 
+<p align="center">
+  <img src="docs/source/_static/intro.gif" alt="SkyPilot">
+</p>
 
 Current supported infra: Kubernetes, AWS, GCP, Azure, OCI, Lambda Cloud, Fluidstack,
 RunPod, Cudo, Digital Ocean, Paperspace, Cloudflare, Samsung, IBM, Vast.ai,
@@ -300,10 +313,6 @@ SkyPilot then performs the heavy-lifting for you, including:
 4. Run the task's `setup` commands to prepare the VM for running the task
 5. Run the task's `run` commands
 
-<p align="center">
-  <img src="https://i.imgur.com/TgamzZ2.gif" alt="SkyPilot Demo"/>
-</p>
-
 
 See [Quickstart](https://docs.skypilot.co/en/latest/getting-started/quickstart.html) to get started with SkyPilot.
 

{skypilot_nightly-1.0.0.dev20250320 → skypilot_nightly-1.0.0.dev20250322}/README.md

@@ -61,8 +61,8 @@ SkyPilot **unifies multiple clusters, clouds, and hardware**:
 
 SkyPilot **cuts your cloud costs & maximizes GPU availability**:
 * Autostop: automatic cleanup of idle resources
-* [Managed Spot](https://docs.skypilot.co/en/latest/examples/managed-jobs.html): 3-6x cost savings using spot instances, with preemption auto-recovery
-* Optimizer: auto-selects the cheapest & most available infra
+* [Spot instance support](https://docs.skypilot.co/en/latest/examples/managed-jobs.html#running-on-spot-instances): 3-6x cost savings, with preemption auto-recovery
+* Intelligent scheduling: automatically run on the cheapest & most available infra
 
 SkyPilot supports your existing GPU, TPU, and CPU workloads, with no code changes.
 
@@ -77,6 +77,9 @@ To get the latest features and fixes, use the nightly build or [install from sou
 pip install "skypilot-nightly[kubernetes,aws,gcp,azure,oci,lambda,runpod,fluidstack,paperspace,cudo,ibm,scp,nebius]"
 ```
 
+<p align="center">
+  <img src="docs/source/_static/intro.gif" alt="SkyPilot">
+</p>
 
 Current supported infra: Kubernetes, AWS, GCP, Azure, OCI, Lambda Cloud, Fluidstack,
 RunPod, Cudo, Digital Ocean, Paperspace, Cloudflare, Samsung, IBM, Vast.ai,
@@ -142,10 +145,6 @@ SkyPilot then performs the heavy-lifting for you, including:
 4. Run the task's `setup` commands to prepare the VM for running the task
 5. Run the task's `run` commands
 
-<p align="center">
-  <img src="https://i.imgur.com/TgamzZ2.gif" alt="SkyPilot Demo"/>
-</p>
-
 
 See [Quickstart](https://docs.skypilot.co/en/latest/getting-started/quickstart.html) to get started with SkyPilot.
 

{skypilot_nightly-1.0.0.dev20250320 → skypilot_nightly-1.0.0.dev20250322}/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 = 'a480f342522afcd17a3b30a20086f28333ddb7b5'
+_SKYPILOT_COMMIT_SHA = '139a09e3445956740743049d352cd1cb6d202479'
 
 
 def _get_git_commit():
@@ -35,7 +35,7 @@ def _get_git_commit():
 
 
 __commit__ = _get_git_commit()
-__version__ = '1.0.0.dev20250320'
+__version__ = '1.0.0.dev20250322'
 __root_dir__ = os.path.dirname(os.path.abspath(__file__))
 
 

{skypilot_nightly-1.0.0.dev20250320 → skypilot_nightly-1.0.0.dev20250322}/sky/adaptors/cloudflare.py

@@ -6,7 +6,9 @@ import os
 import threading
 from typing import Dict, Optional, Tuple
 
+from sky import exceptions
 from sky.adaptors import common
+from sky.clouds import cloud
 from sky.utils import annotations
 from sky.utils import ux_utils
 
@@ -23,7 +25,6 @@ R2_CREDENTIALS_PATH = '~/.cloudflare/r2.credentials'
 R2_PROFILE_NAME = 'r2'
 _INDENT_PREFIX = '    '
 NAME = 'Cloudflare'
-SKY_CHECK_NAME = 'Cloudflare (for R2 object store)'
 
 
 @contextlib.contextmanager
@@ -130,8 +131,8 @@ def client(service_name: str, region):
 @common.load_lazy_modules(_LAZY_MODULES)
 def botocore_exceptions():
     """AWS botocore exception."""
-    from botocore import exceptions
-    return exceptions
+    from botocore import exceptions as boto_exceptions
+    return boto_exceptions
 
 
 def create_endpoint():
@@ -148,8 +149,18 @@ def create_endpoint():
     return endpoint
 
 
-def check_credentials() -> Tuple[bool, Optional[str]]:
-    return check_storage_credentials()
+def check_credentials(
+        cloud_capability: cloud.CloudCapability) -> Tuple[bool, Optional[str]]:
+    if cloud_capability == cloud.CloudCapability.COMPUTE:
+        # for backward compatibility,
+        # we check storage credentials for compute.
+        # TODO(seungjin): properly return not supported error for compute.
+        return check_storage_credentials()
+    elif cloud_capability == cloud.CloudCapability.STORAGE:
+        return check_storage_credentials()
+    else:
+        raise exceptions.NotSupportedError(
+            f'{NAME} does not support {cloud_capability}.')
 
 
 def check_storage_credentials() -> Tuple[bool, Optional[str]]:

{skypilot_nightly-1.0.0.dev20250320 → skypilot_nightly-1.0.0.dev20250322}/sky/adaptors/kubernetes.py

@@ -79,10 +79,11 @@ def _load_config(context: Optional[str] = None):
                     '    If you were running a local Kubernetes '
                     'cluster, run `sky local up` to start the cluster.')
             else:
+                kubeconfig_path = os.environ.get('KUBECONFIG', '~/.kube/config')
                 err_str = (
                     f'Failed to load Kubernetes configuration for {context!r}. '
                     'Please check if your kubeconfig file exists at '
-                    f'~/.kube/config and is valid.\n{suffix}')
+                    f'{kubeconfig_path} and is valid.\n{suffix}')
             err_str += '\nTo disable Kubernetes for SkyPilot: run `sky check`.'
             with ux_utils.print_exception_no_traceback():
                 raise ValueError(err_str) from None

skypilot_nightly-1.0.0.dev20250322/sky/adaptors/nebius.py

@@ -0,0 +1,232 @@
+"""Nebius cloud adaptor."""
+import os
+import threading
+from typing import Optional
+
+from sky.adaptors import common
+from sky.utils import annotations
+from sky.utils import ux_utils
+
+NEBIUS_TENANT_ID_FILENAME = 'NEBIUS_TENANT_ID.txt'
+NEBIUS_IAM_TOKEN_FILENAME = 'NEBIUS_IAM_TOKEN.txt'
+NEBIUS_PROJECT_ID_FILENAME = 'NEBIUS_PROJECT_ID.txt'
+NEBIUS_CREDENTIALS_FILENAME = 'credentials.json'
+NEBIUS_TENANT_ID_PATH = '~/.nebius/' + NEBIUS_TENANT_ID_FILENAME
+NEBIUS_IAM_TOKEN_PATH = '~/.nebius/' + NEBIUS_IAM_TOKEN_FILENAME
+NEBIUS_PROJECT_ID_PATH = '~/.nebius/' + NEBIUS_PROJECT_ID_FILENAME
+NEBIUS_CREDENTIALS_PATH = '~/.nebius/' + NEBIUS_CREDENTIALS_FILENAME
+
+DEFAULT_REGION = 'eu-north1'
+
+NEBIUS_PROFILE_NAME = 'nebius'
+
+MAX_RETRIES_TO_DISK_CREATE = 120
+MAX_RETRIES_TO_INSTANCE_STOP = 120
+MAX_RETRIES_TO_INSTANCE_START = 120
+MAX_RETRIES_TO_INSTANCE_READY = 240
+
+MAX_RETRIES_TO_DISK_DELETE = 120
+MAX_RETRIES_TO_INSTANCE_WAIT = 120  # Maximum number of retries
+
+POLL_INTERVAL = 5
+
+_iam_token = None
+_sdk = None
+_tenant_id = None
+_project_id = None
+
+_IMPORT_ERROR_MESSAGE = ('Failed to import dependencies for Nebius AI Cloud.'
+                         'Try pip install "skypilot[nebius]"')
+
+nebius = common.LazyImport(
+    'nebius',
+    import_error_message=_IMPORT_ERROR_MESSAGE,
+    # https://github.com/grpc/grpc/issues/37642 to avoid spam in console
+    set_loggers=lambda: os.environ.update({'GRPC_VERBOSITY': 'NONE'}))
+boto3 = common.LazyImport('boto3', import_error_message=_IMPORT_ERROR_MESSAGE)
+botocore = common.LazyImport('botocore',
+                             import_error_message=_IMPORT_ERROR_MESSAGE)
+
+_LAZY_MODULES = (boto3, botocore, nebius)
+_session_creation_lock = threading.RLock()
+_INDENT_PREFIX = '    '
+NAME = 'Nebius'
+SKY_CHECK_NAME = 'Nebius (for Nebius Object Storae)'
+
+
+def request_error():
+    return nebius.aio.service_error.RequestError
+
+
+def compute():
+    # pylint: disable=import-outside-toplevel
+    from nebius.api.nebius.compute import v1 as compute_v1
+    return compute_v1
+
+
+def iam():
+    # pylint: disable=import-outside-toplevel
+    from nebius.api.nebius.iam import v1 as iam_v1
+    return iam_v1
+
+
+def nebius_common():
+    # pylint: disable=import-outside-toplevel
+    from nebius.api.nebius.common import v1 as common_v1
+    return common_v1
+
+
+def vpc():
+    # pylint: disable=import-outside-toplevel
+    from nebius.api.nebius.vpc import v1 as vpc_v1
+    return vpc_v1
+
+
+def get_iam_token():
+    global _iam_token
+    if _iam_token is None:
+        try:
+            with open(os.path.expanduser(NEBIUS_IAM_TOKEN_PATH),
+                      encoding='utf-8') as file:
+                _iam_token = file.read().strip()
+        except FileNotFoundError:
+            return None
+    return _iam_token
+
+
+def is_token_or_cred_file_exist():
+    return (os.path.exists(os.path.expanduser(NEBIUS_IAM_TOKEN_PATH)) or
+            os.path.exists(os.path.expanduser(NEBIUS_CREDENTIALS_PATH)))
+
+
+def get_project_id():
+    global _project_id
+    if _project_id is None:
+        try:
+            with open(os.path.expanduser(NEBIUS_PROJECT_ID_PATH),
+                      encoding='utf-8') as file:
+                _project_id = file.read().strip()
+        except FileNotFoundError:
+            return None
+    return _project_id
+
+
+def get_tenant_id():
+    global _tenant_id
+    if _tenant_id is None:
+        try:
+            with open(os.path.expanduser(NEBIUS_TENANT_ID_PATH),
+                      encoding='utf-8') as file:
+                _tenant_id = file.read().strip()
+        except FileNotFoundError:
+            return None
+    return _tenant_id
+
+
+def sdk():
+    global _sdk
+    if _sdk is None:
+        if get_iam_token() is not None:
+            _sdk = nebius.sdk.SDK(credentials=get_iam_token())
+            return _sdk
+        _sdk = nebius.sdk.SDK(
+            credentials_file_name=os.path.expanduser(NEBIUS_CREDENTIALS_PATH))
+    return _sdk
+
+
+def get_nebius_credentials(boto3_session):
+    """Gets the Nebius credentials from the boto3 session object.
+
+    Args:
+        boto3_session: The boto3 session object.
+    Returns:
+        botocore.credentials.ReadOnlyCredentials object with the R2 credentials.
+    """
+    nebius_credentials = boto3_session.get_credentials()
+    if nebius_credentials is None:
+        with ux_utils.print_exception_no_traceback():
+            raise ValueError('Nebius credentials not found. Run '
+                             '`sky check` to verify credentials are '
+                             'correctly set up.')
+    return nebius_credentials.get_frozen_credentials()
+
+
+# lru_cache() is thread-safe and it will return the same session object
+# for different threads.
+# Reference: https://docs.python.org/3/library/functools.html#functools.lru_cache # pylint: disable=line-too-long
+@annotations.lru_cache(scope='global')
+def session():
+    """Create an AWS session."""
+    # Creating the session object is not thread-safe for boto3,
+    # so we add a reentrant lock to synchronize the session creation.
+    # Reference: https://github.com/boto/boto3/issues/1592
+    # However, the session object itself is thread-safe, so we are
+    # able to use lru_cache() to cache the session object.
+    with _session_creation_lock:
+        session_ = boto3.session.Session(profile_name=NEBIUS_PROFILE_NAME)
+    return session_
+
+
+@annotations.lru_cache(scope='global')
+def resource(resource_name: str, region: str = DEFAULT_REGION, **kwargs):
+    """Create a Nebius resource.
+
+    Args:
+        resource_name: Nebius resource name (e.g., 's3').
+        kwargs: Other options.
+    """
+    # Need to use the resource retrieved from the per-thread session
+    # to avoid thread-safety issues (Directly creating the client
+    # with boto3.resource() is not thread-safe).
+    # Reference: https://stackoverflow.com/a/59635814
+
+    session_ = session()
+    nebius_credentials = get_nebius_credentials(session_)
+    endpoint = create_endpoint(region)
+
+    return session_.resource(
+        resource_name,
+        endpoint_url=endpoint,
+        aws_access_key_id=nebius_credentials.access_key,
+        aws_secret_access_key=nebius_credentials.secret_key,
+        region_name=region,
+        **kwargs)
+
+
+@annotations.lru_cache(scope='global')
+def client(service_name: str, region):
+    """Create an Nebius client of a certain service.
+
+    Args:
+        service_name: Nebius service name (e.g., 's3').
+        kwargs: Other options.
+    """
+    # Need to use the client retrieved from the per-thread session
+    # to avoid thread-safety issues (Directly creating the client
+    # with boto3.client() is not thread-safe).
+    # Reference: https://stackoverflow.com/a/59635814
+
+    session_ = session()
+    nebius_credentials = get_nebius_credentials(session_)
+    endpoint = create_endpoint(region)
+
+    return session_.client(service_name,
+                           endpoint_url=endpoint,
+                           aws_access_key_id=nebius_credentials.access_key,
+                           aws_secret_access_key=nebius_credentials.secret_key,
+                           region_name=region)
+
+
+@common.load_lazy_modules(_LAZY_MODULES)
+def botocore_exceptions():
+    """AWS botocore exception."""
+    # pylint: disable=import-outside-toplevel
+    from botocore import exceptions
+    return exceptions
+
+
+def create_endpoint(region: Optional[str] = DEFAULT_REGION) -> str:
+    """Reads accountid necessary to interact with Nebius Object Storage"""
+    if region is None:
+        region = DEFAULT_REGION
+    return f'https://storage.{region}.nebius.cloud:443'

{skypilot_nightly-1.0.0.dev20250320 → skypilot_nightly-1.0.0.dev20250322}/sky/backends/cloud_vm_ray_backend.py

@@ -38,6 +38,7 @@ from sky import sky_logging
 from sky import task as task_lib
 from sky.backends import backend_utils
 from sky.backends import wheel_utils
+from sky.clouds import cloud as sky_cloud
 from sky.clouds import service_catalog
 from sky.clouds.utils import gcp_utils
 from sky.data import data_utils
@@ -1981,7 +1982,8 @@ class RetryingVmProvisioner(object):
         # is running. Here we check the enabled clouds and expiring credentials
         # and raise a warning to the user.
         if task.is_controller_task():
-            enabled_clouds = sky_check.get_cached_enabled_clouds_or_refresh()
+            enabled_clouds = sky_check.get_cached_enabled_clouds_or_refresh(
+                sky_cloud.CloudCapability.COMPUTE)
             expirable_clouds = backend_utils.get_expirable_clouds(
                 enabled_clouds)
 

{skypilot_nightly-1.0.0.dev20250320 → skypilot_nightly-1.0.0.dev20250322}/sky/benchmark/benchmark_utils.py

@@ -172,8 +172,9 @@ def _create_benchmark_bucket() -> Tuple[str, str]:
     bucket_name = f'sky-bench-{uuid.uuid4().hex[:4]}-{getpass.getuser()}'
 
     # Select the bucket type.
-    enabled_clouds = storage_lib.get_cached_enabled_storage_clouds_or_refresh(
-        raise_if_no_cloud_access=True)
+    enabled_clouds = (
+        storage_lib.get_cached_enabled_storage_cloud_names_or_refresh(
+            raise_if_no_cloud_access=True))
     # Sky Benchmark only supports S3 (see _download_remote_dir and
     # _delete_remote_dir).
     enabled_clouds = [