skypilot-nightly 1.0.0.dev20250410__tar.gz → 1.0.0.dev20250412__tar.gz

{skypilot_nightly-1.0.0.dev20250410/skypilot_nightly.egg-info → skypilot_nightly-1.0.0.dev20250412}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: skypilot-nightly
-Version: 1.0.0.dev20250410
+Version: 1.0.0.dev20250412
 Summary: SkyPilot: An intercloud broker for the clouds
 Author: SkyPilot Team
 License: Apache 2.0
@@ -88,6 +88,7 @@ Requires-Dist: oci; extra == "oci"
 Requires-Dist: ray[default]!=2.6.0,>=2.2.0; extra == "oci"
 Provides-Extra: kubernetes
 Requires-Dist: kubernetes!=32.0.0,>=20.0.0; extra == "kubernetes"
+Requires-Dist: websockets; extra == "kubernetes"
 Provides-Extra: remote
 Requires-Dist: grpcio!=1.48.0,>=1.32.0; python_version < "3.10" and extra == "remote"
 Requires-Dist: grpcio!=1.48.0,>=1.42.0; python_version >= "3.10" and extra == "remote"
@@ -145,6 +146,7 @@ Requires-Dist: ray[default]!=2.6.0,>=2.2.0; extra == "all"
 Requires-Dist: oci; extra == "all"
 Requires-Dist: ray[default]!=2.6.0,>=2.2.0; extra == "all"
 Requires-Dist: kubernetes!=32.0.0,>=20.0.0; extra == "all"
+Requires-Dist: websockets; extra == "all"
 Requires-Dist: grpcio!=1.48.0,>=1.32.0; python_version < "3.10" and extra == "all"
 Requires-Dist: grpcio!=1.48.0,>=1.42.0; python_version >= "3.10" and extra == "all"
 Requires-Dist: protobuf!=3.19.5,>=3.15.3; extra == "all"

{skypilot_nightly-1.0.0.dev20250410 → skypilot_nightly-1.0.0.dev20250412}/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 = '022a5c3ffe258f365764b03cb20fac70934f5a60'
+_SKYPILOT_COMMIT_SHA = 'ac3cbfe07c8c1129480d4facd683f4844ff6e1f7'
 
 
 def _get_git_commit():
@@ -35,7 +35,7 @@ def _get_git_commit():
 
 
 __commit__ = _get_git_commit()
-__version__ = '1.0.0.dev20250410'
+__version__ = '1.0.0.dev20250412'
 __root_dir__ = os.path.dirname(os.path.abspath(__file__))
 
 

{skypilot_nightly-1.0.0.dev20250410 → skypilot_nightly-1.0.0.dev20250412}/sky/adaptors/oci.py

@@ -13,7 +13,7 @@ from sky.clouds.utils import oci_utils
 # effect.
 logging.getLogger('oci.circuit_breaker').setLevel(logging.WARNING)
 
-CONFIG_PATH = '~/.oci/config'
+OCI_CONFIG_PATH = '~/.oci/config'
 ENV_VAR_OCI_CONFIG = 'OCI_CONFIG'
 
 oci = common.LazyImport(
@@ -23,7 +23,7 @@ oci = common.LazyImport(
 
 
 def get_config_file() -> str:
-    conf_file_path = CONFIG_PATH
+    conf_file_path = OCI_CONFIG_PATH
     config_path_via_env_var = os.environ.get(ENV_VAR_OCI_CONFIG)
     if config_path_via_env_var is not None:
         conf_file_path = config_path_via_env_var

{skypilot_nightly-1.0.0.dev20250410 → skypilot_nightly-1.0.0.dev20250412}/sky/authentication.py

@@ -382,10 +382,10 @@ def setup_kubernetes_authentication(config: Dict[str, Any]) -> Dict[str, Any]:
         network_mode = kubernetes_enums.KubernetesNetworkingMode.from_str(
             network_mode_str)
     except ValueError as e:
-        # Add message saying "Please check: ~/.sky/config.yaml" to the error
+        # Add message saying "Please check: ~/.sky/skyconfig.yaml" to the error
         # message.
         with ux_utils.print_exception_no_traceback():
-            raise ValueError(str(e) + ' Please check: ~/.sky/config.yaml.') \
+            raise ValueError(str(e) + ' Please check: ~/.sky/skyconfig.yaml.') \
                 from None
     _, public_key_path = get_or_generate_keys()
 

{skypilot_nightly-1.0.0.dev20250410 → skypilot_nightly-1.0.0.dev20250412}/sky/backends/backend_utils.py

@@ -682,7 +682,7 @@ def write_cluster_config(
             ssh_proxy_command = ssh_proxy_command_config[region_name]
     logger.debug(f'Using ssh_proxy_command: {ssh_proxy_command!r}')
 
-    # User-supplied global instance tags from ~/.sky/config.yaml.
+    # User-supplied global instance tags from ~/.sky/skyconfig.yaml.
     labels = skypilot_config.get_nested((str(cloud).lower(), 'labels'), {})
     # labels is a dict, which is guaranteed by the type check in
     # schemas.py

{skypilot_nightly-1.0.0.dev20250410 → skypilot_nightly-1.0.0.dev20250412}/sky/backends/cloud_vm_ray_backend.py

@@ -1473,7 +1473,7 @@ class RetryingVmProvisioner(object):
                     f'invalid cloud credentials: '
                     f'{common_utils.format_exception(e)}')
             except exceptions.InvalidCloudConfigs as e:
-                # Failed due to invalid user configs in ~/.sky/config.yaml.
+                # Failed due to invalid user configs in ~/.sky/skyconfig.yaml.
                 logger.warning(f'{common_utils.format_exception(e)}')
                 # We should block the entire cloud if the user config is
                 # invalid.
@@ -2065,10 +2065,10 @@ class RetryingVmProvisioner(object):
                                (clouds.Kubernetes, clouds.RunPod)) and
                         controller_utils.Controllers.from_name(cluster_name)
                         is not None):
-                    assert (clouds.CloudImplementationFeatures.STOP
+                    assert (clouds.CloudImplementationFeatures.AUTOSTOP
                             in requested_features), requested_features
                     requested_features.remove(
-                        clouds.CloudImplementationFeatures.STOP)
+                        clouds.CloudImplementationFeatures.AUTOSTOP)
 
                 # Skip if to_provision.cloud does not support requested features
                 to_provision.cloud.check_features_are_supported(

{skypilot_nightly-1.0.0.dev20250410 → skypilot_nightly-1.0.0.dev20250412}/sky/check.py

@@ -142,7 +142,7 @@ def check_capabilities(
     if disallowed_cloud_names:
         disallowed_clouds_hint = (
             '\nNote: The following clouds were disabled because they were not '
-            'included in allowed_clouds in ~/.sky/config.yaml: '
+            'included in allowed_clouds in ~/.sky/skyconfig.yaml: '
             f'{", ".join([c for c in disallowed_cloud_names])}')
     if not all_enabled_clouds:
         echo(

{skypilot_nightly-1.0.0.dev20250410 → skypilot_nightly-1.0.0.dev20250412}/sky/cli.py

@@ -35,7 +35,8 @@ import sys
 import textwrap
 import traceback
 import typing
-from typing import Any, Callable, Dict, Generator, List, Optional, Tuple, Union
+from typing import (Any, Callable, Dict, Generator, List, Optional, Set, Tuple,
+                    Union)
 
 import click
 import colorama
@@ -134,49 +135,51 @@ def _get_cluster_records_and_set_ssh_config(
     # Update the SSH config for all clusters
     for record in cluster_records:
         handle = record['handle']
-        # During the failover, even though a cluster does not exist, the handle
-        # can still exist in the record, and we check for credentials to avoid
-        # updating the SSH config for non-existent clusters.
-        if (handle is not None and handle.cached_external_ips is not None and
-                'credentials' in record):
-            credentials = record['credentials']
-            if isinstance(handle.launched_resources.cloud, clouds.Kubernetes):
-                # Replace the proxy command to proxy through the SkyPilot API
-                # server with websocket.
-                key_path = (
-                    cluster_utils.SSHConfigHelper.generate_local_key_file(
-                        handle.cluster_name, credentials))
-                # Instead of directly use websocket_proxy.py, we add an
-                # additional proxy, so that ssh can use the head pod in the
-                # cluster to jump to worker pods.
-                proxy_command = (
-                    f'ssh -tt -i {key_path} '
-                    '-o StrictHostKeyChecking=no '
-                    '-o UserKnownHostsFile=/dev/null '
-                    '-o IdentitiesOnly=yes '
-                    '-W %h:%p '
-                    f'{handle.ssh_user}@127.0.0.1 '
-                    '-o ProxyCommand='
-                    # TODO(zhwu): write the template to a temp file, don't use
-                    # the one in skypilot repo, to avoid changing the file when
-                    # updating skypilot.
-                    f'\'{sys.executable} {sky.__root_dir__}/templates/'
-                    f'websocket_proxy.py '
-                    f'{server_common.get_server_url().split("://")[1]} '
-                    f'{handle.cluster_name}\'')
-                credentials['ssh_proxy_command'] = proxy_command
-            cluster_utils.SSHConfigHelper.add_cluster(
-                handle.cluster_name,
-                handle.cached_external_ips,
-                credentials,
-                handle.cached_external_ssh_ports,
-                handle.docker_user,
-                handle.ssh_user,
-            )
-        else:
+
+        if not (handle is not None and handle.cached_external_ips is not None
+                and 'credentials' in record):
             # If the cluster is not UP or does not have credentials available,
             # we need to remove the cluster from the SSH config.
             cluster_utils.SSHConfigHelper.remove_cluster(record['name'])
+            continue
+
+        # During the failover, even though a cluster does not exist, the handle
+        # can still exist in the record, and we check for credentials to avoid
+        # updating the SSH config for non-existent clusters.
+        credentials = record['credentials']
+        if isinstance(handle.launched_resources.cloud, clouds.Kubernetes):
+            # Replace the proxy command to proxy through the SkyPilot API
+            # server with websocket.
+            key_path = (cluster_utils.SSHConfigHelper.generate_local_key_file(
+                handle.cluster_name, credentials))
+            # Instead of directly use websocket_proxy.py, we add an
+            # additional proxy, so that ssh can use the head pod in the
+            # cluster to jump to worker pods.
+            proxy_command = (
+                f'ssh -tt -i {key_path} '
+                '-o StrictHostKeyChecking=no '
+                '-o UserKnownHostsFile=/dev/null '
+                '-o IdentitiesOnly=yes '
+                '-W %h:%p '
+                f'{handle.ssh_user}@127.0.0.1 '
+                '-o ProxyCommand='
+                # TODO(zhwu): write the template to a temp file, don't use
+                # the one in skypilot repo, to avoid changing the file when
+                # updating skypilot.
+                f'\'{sys.executable} {sky.__root_dir__}/templates/'
+                f'websocket_proxy.py '
+                f'{server_common.get_server_url()} '
+                f'{handle.cluster_name}\'')
+            credentials['ssh_proxy_command'] = proxy_command
+
+        cluster_utils.SSHConfigHelper.add_cluster(
+            handle.cluster_name,
+            handle.cached_external_ips,
+            credentials,
+            handle.cached_external_ssh_ports,
+            handle.docker_user,
+            handle.ssh_user,
+        )
 
     # Clean up SSH configs for clusters that do not exist.
     #
@@ -186,14 +189,15 @@ def _get_cluster_records_and_set_ssh_config(
     # removing clusters, because SkyPilot has no idea whether to remove
     # ssh config of a cluster from another user.
     clusters_exists = set(record['name'] for record in cluster_records)
+    clusters_to_remove: Set[str] = set()
     if clusters is not None:
-        for cluster in clusters:
-            if cluster not in clusters_exists:
-                cluster_utils.SSHConfigHelper.remove_cluster(cluster)
+        clusters_to_remove = set(clusters) - clusters_exists
     elif all_users:
-        for cluster_name in cluster_utils.SSHConfigHelper.list_cluster_names():
-            if cluster_name not in clusters_exists:
-                cluster_utils.SSHConfigHelper.remove_cluster(cluster_name)
+        clusters_to_remove = set(cluster_utils.SSHConfigHelper.
+                                 list_cluster_names()) - clusters_exists
+
+    for cluster_name in clusters_to_remove:
+        cluster_utils.SSHConfigHelper.remove_cluster(cluster_name)
 
     return cluster_records
 

{skypilot_nightly-1.0.0.dev20250410 → skypilot_nightly-1.0.0.dev20250412}/sky/client/cli.py

@@ -35,7 +35,8 @@ import sys
 import textwrap
 import traceback
 import typing
-from typing import Any, Callable, Dict, Generator, List, Optional, Tuple, Union
+from typing import (Any, Callable, Dict, Generator, List, Optional, Set, Tuple,
+                    Union)
 
 import click
 import colorama
@@ -134,49 +135,51 @@ def _get_cluster_records_and_set_ssh_config(
     # Update the SSH config for all clusters
     for record in cluster_records:
         handle = record['handle']
-        # During the failover, even though a cluster does not exist, the handle
-        # can still exist in the record, and we check for credentials to avoid
-        # updating the SSH config for non-existent clusters.
-        if (handle is not None and handle.cached_external_ips is not None and
-                'credentials' in record):
-            credentials = record['credentials']
-            if isinstance(handle.launched_resources.cloud, clouds.Kubernetes):
-                # Replace the proxy command to proxy through the SkyPilot API
-                # server with websocket.
-                key_path = (
-                    cluster_utils.SSHConfigHelper.generate_local_key_file(
-                        handle.cluster_name, credentials))
-                # Instead of directly use websocket_proxy.py, we add an
-                # additional proxy, so that ssh can use the head pod in the
-                # cluster to jump to worker pods.
-                proxy_command = (
-                    f'ssh -tt -i {key_path} '
-                    '-o StrictHostKeyChecking=no '
-                    '-o UserKnownHostsFile=/dev/null '
-                    '-o IdentitiesOnly=yes '
-                    '-W %h:%p '
-                    f'{handle.ssh_user}@127.0.0.1 '
-                    '-o ProxyCommand='
-                    # TODO(zhwu): write the template to a temp file, don't use
-                    # the one in skypilot repo, to avoid changing the file when
-                    # updating skypilot.
-                    f'\'{sys.executable} {sky.__root_dir__}/templates/'
-                    f'websocket_proxy.py '
-                    f'{server_common.get_server_url().split("://")[1]} '
-                    f'{handle.cluster_name}\'')
-                credentials['ssh_proxy_command'] = proxy_command
-            cluster_utils.SSHConfigHelper.add_cluster(
-                handle.cluster_name,
-                handle.cached_external_ips,
-                credentials,
-                handle.cached_external_ssh_ports,
-                handle.docker_user,
-                handle.ssh_user,
-            )
-        else:
+
+        if not (handle is not None and handle.cached_external_ips is not None
+                and 'credentials' in record):
             # If the cluster is not UP or does not have credentials available,
             # we need to remove the cluster from the SSH config.
             cluster_utils.SSHConfigHelper.remove_cluster(record['name'])
+            continue
+
+        # During the failover, even though a cluster does not exist, the handle
+        # can still exist in the record, and we check for credentials to avoid
+        # updating the SSH config for non-existent clusters.
+        credentials = record['credentials']
+        if isinstance(handle.launched_resources.cloud, clouds.Kubernetes):
+            # Replace the proxy command to proxy through the SkyPilot API
+            # server with websocket.
+            key_path = (cluster_utils.SSHConfigHelper.generate_local_key_file(
+                handle.cluster_name, credentials))
+            # Instead of directly use websocket_proxy.py, we add an
+            # additional proxy, so that ssh can use the head pod in the
+            # cluster to jump to worker pods.
+            proxy_command = (
+                f'ssh -tt -i {key_path} '
+                '-o StrictHostKeyChecking=no '
+                '-o UserKnownHostsFile=/dev/null '
+                '-o IdentitiesOnly=yes '
+                '-W %h:%p '
+                f'{handle.ssh_user}@127.0.0.1 '
+                '-o ProxyCommand='
+                # TODO(zhwu): write the template to a temp file, don't use
+                # the one in skypilot repo, to avoid changing the file when
+                # updating skypilot.
+                f'\'{sys.executable} {sky.__root_dir__}/templates/'
+                f'websocket_proxy.py '
+                f'{server_common.get_server_url()} '
+                f'{handle.cluster_name}\'')
+            credentials['ssh_proxy_command'] = proxy_command
+
+        cluster_utils.SSHConfigHelper.add_cluster(
+            handle.cluster_name,
+            handle.cached_external_ips,
+            credentials,
+            handle.cached_external_ssh_ports,
+            handle.docker_user,
+            handle.ssh_user,
+        )
 
     # Clean up SSH configs for clusters that do not exist.
     #
@@ -186,14 +189,15 @@ def _get_cluster_records_and_set_ssh_config(
     # removing clusters, because SkyPilot has no idea whether to remove
     # ssh config of a cluster from another user.
     clusters_exists = set(record['name'] for record in cluster_records)
+    clusters_to_remove: Set[str] = set()
     if clusters is not None:
-        for cluster in clusters:
-            if cluster not in clusters_exists:
-                cluster_utils.SSHConfigHelper.remove_cluster(cluster)
+        clusters_to_remove = set(clusters) - clusters_exists
     elif all_users:
-        for cluster_name in cluster_utils.SSHConfigHelper.list_cluster_names():
-            if cluster_name not in clusters_exists:
-                cluster_utils.SSHConfigHelper.remove_cluster(cluster_name)
+        clusters_to_remove = set(cluster_utils.SSHConfigHelper.
+                                 list_cluster_names()) - clusters_exists
+
+    for cluster_name in clusters_to_remove:
+        cluster_utils.SSHConfigHelper.remove_cluster(cluster_name)
 
     return cluster_records
 

{skypilot_nightly-1.0.0.dev20250410 → skypilot_nightly-1.0.0.dev20250412}/sky/client/common.py

@@ -75,7 +75,8 @@ def download_logs_from_api_server(
     body = payloads.DownloadBody(folder_paths=list(paths_on_api_server),)
     response = requests.post(f'{server_common.get_server_url()}/download',
                              json=json.loads(body.model_dump_json()),
-                             stream=True)
+                             stream=True,
+                             cookies=server_common.get_api_cookie_jar())
     if response.status_code == 200:
         remote_home_path = response.headers.get('X-Home-Path')
         assert remote_home_path is not None, response.headers
@@ -176,7 +177,8 @@ def _upload_chunk_with_retry(params: UploadChunkParams) -> None:
                 },
                 content=FileChunkIterator(f, _UPLOAD_CHUNK_BYTES,
                                           params.chunk_index),
-                headers={'Content-Type': 'application/octet-stream'})
+                headers={'Content-Type': 'application/octet-stream'},
+                cookies=server_common.get_api_cookie_jar())
             if response.status_code == 200:
                 data = response.json()
                 status = data.get('status')