skypilot-nightly 1.0.0.dev2024053101__py3-none-any.whl → 1.0.0.dev2025022801__py3-none-any.whl

sky/utils/command_runner.pyi

@@ -16,7 +16,8 @@ from sky.utils import subprocess_utils as subprocess_utils
 
 GIT_EXCLUDE: str
 RSYNC_DISPLAY_OPTION: str
-RSYNC_FILTER_OPTION: str
+RSYNC_FILTER_GITIGNORE: str
+RSYNC_FILTER_SKYIGNORE: str
 RSYNC_EXCLUDE_OPTION: str
 ALIAS_SUDO_TO_EMPTY_FOR_ROOT_CMD: str
 
@@ -102,7 +103,7 @@ class CommandRunner:
               up: bool,
               log_path: str = ...,
               stream_logs: bool = ...,
-              max_retry: int = 1) -> None:
+              max_retry: int = ...) -> None:
         ...
 
     @classmethod
@@ -114,6 +115,9 @@ class CommandRunner:
     def check_connection(self) -> bool:
         ...
 
+    def close_cached_connection(self) -> None:
+        ...
+
 
 class SSHCommandRunner(CommandRunner):
     ip: str
@@ -193,5 +197,75 @@ class SSHCommandRunner(CommandRunner):
               up: bool,
               log_path: str = ...,
               stream_logs: bool = ...,
-              max_retry: int = 1) -> None:
+              max_retry: int = ...) -> None:
+        ...
+
+
+class KubernetesCommandRunner(CommandRunner):
+
+    def __init__(
+        self,
+        node: Tuple[Tuple[str, Optional[str]], str],
+    ) -> None:
+        ...
+
+    @typing.overload
+    def run(self,
+            cmd: Union[str, List[str]],
+            *,
+            port_forward: Optional[List[int]] = ...,
+            require_outputs: Literal[False] = ...,
+            log_path: str = ...,
+            process_stream: bool = ...,
+            stream_logs: bool = ...,
+            ssh_mode: SshMode = ...,
+            separate_stderr: bool = ...,
+            connect_timeout: Optional[int] = ...,
+            source_bashrc: bool = ...,
+            skip_lines: int = ...,
+            **kwargs) -> int:
+        ...
+
+    @typing.overload
+    def run(self,
+            cmd: Union[str, List[str]],
+            *,
+            port_forward: Optional[List[int]] = ...,
+            require_outputs: Literal[True],
+            log_path: str = ...,
+            process_stream: bool = ...,
+            stream_logs: bool = ...,
+            ssh_mode: SshMode = ...,
+            separate_stderr: bool = ...,
+            connect_timeout: Optional[int] = ...,
+            source_bashrc: bool = ...,
+            skip_lines: int = ...,
+            **kwargs) -> Tuple[int, str, str]:
+        ...
+
+    @typing.overload
+    def run(self,
+            cmd: Union[str, List[str]],
+            *,
+            port_forward: Optional[List[int]] = ...,
+            require_outputs: bool = ...,
+            log_path: str = ...,
+            process_stream: bool = ...,
+            stream_logs: bool = ...,
+            ssh_mode: SshMode = ...,
+            separate_stderr: bool = ...,
+            connect_timeout: Optional[int] = ...,
+            source_bashrc: bool = ...,
+            skip_lines: int = ...,
+            **kwargs) -> Union[Tuple[int, str, str], int]:
+        ...
+
+    def rsync(self,
+              source: str,
+              target: str,
+              *,
+              up: bool,
+              log_path: str = ...,
+              stream_logs: bool = ...,
+              max_retry: int = ...) -> None:
         ...

sky/utils/common.py

@@ -0,0 +1,54 @@
+"""Common enumerators and classes."""
+
+import contextlib
+import enum
+import os
+from typing import Generator
+
+from sky.skylet import constants
+from sky.utils import common_utils
+
+SKY_SERVE_CONTROLLER_PREFIX: str = 'sky-serve-controller-'
+JOB_CONTROLLER_PREFIX: str = 'sky-jobs-controller-'
+# We use the user hash (machine-specific) for the controller name. It will be
+# the same across the whole lifecycle of the server, including:
+# 1. all requests, because this global variable is set once during server
+#    starts.
+# 2. SkyPilot API server restarts, as long as the `~/.sky` folder is persisted
+#    and the env var set during starting the server is the same.
+# This behavior is the same for the local API server (where SERVER_ID is the
+# same as the normal user hash). This ensures backwards-compatibility with jobs
+# controllers from before #4660.
+SERVER_ID = common_utils.get_user_hash()
+SKY_SERVE_CONTROLLER_NAME: str = f'{SKY_SERVE_CONTROLLER_PREFIX}{SERVER_ID}'
+JOB_CONTROLLER_NAME: str = f'{JOB_CONTROLLER_PREFIX}{SERVER_ID}'
+
+
+@contextlib.contextmanager
+def with_server_user_hash() -> Generator[None, None, None]:
+    """Temporarily set the user hash to common.SERVER_ID."""
+    old_env_user_hash = os.getenv(constants.USER_ID_ENV_VAR)
+    os.environ[constants.USER_ID_ENV_VAR] = SERVER_ID
+    try:
+        yield
+    finally:
+        if old_env_user_hash is not None:
+            os.environ[constants.USER_ID_ENV_VAR] = old_env_user_hash
+        else:
+            os.environ.pop(constants.USER_ID_ENV_VAR)
+
+
+class StatusRefreshMode(enum.Enum):
+    """The mode of refreshing the status of a cluster."""
+    NONE = 'NONE'
+    """Do not refresh any clusters."""
+    AUTO = 'AUTO'
+    """Only refresh clusters if their autostop is set or have spot instances."""
+    FORCE = 'FORCE'
+    """Enforce refreshing all clusters."""
+
+
+# Constants: minimize what target?
+class OptimizeTarget(enum.Enum):
+    COST = 0
+    TIME = 1