skypilot-nightly 1.0.0.dev20250303__py3-none-any.whl → 1.0.0.dev20250305__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 = '50bbf9162d3516d5eb4e59e071822a4f3c38622c'
+_SKYPILOT_COMMIT_SHA = '296a22e868b9bdf1faccbe3effbfb858a5a05905'
 
 
 def _get_git_commit():
@@ -35,7 +35,7 @@ def _get_git_commit():
 
 
 __commit__ = _get_git_commit()
-__version__ = '1.0.0.dev20250303'
+__version__ = '1.0.0.dev20250305'
 __root_dir__ = os.path.dirname(os.path.abspath(__file__))
 
 

sky/backends/cloud_vm_ray_backend.py

@@ -3823,6 +3823,10 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
             follow: Whether to follow the logs.
             tail: The number of lines to display from the end of the
                 log file. If 0, print all lines.
+
+        Returns:
+            The exit code of the tail command. Returns code 100 if the job has
+            failed. See exceptions.JobExitCode for possible return codes.
         """
         code = job_lib.JobLibCodeGen.tail_logs(job_id,
                                                managed_job_id=managed_job_id,
@@ -3856,7 +3860,7 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
                               job_id: Optional[int] = None,
                               job_name: Optional[str] = None,
                               controller: bool = False,
-                              follow: bool = True) -> None:
+                              follow: bool = True) -> 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)
         code = managed_jobs.ManagedJobCodeGen.stream_logs(
@@ -3869,13 +3873,17 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
             signal.signal(signal.SIGTSTP, backend_utils.stop_handler)
 
         # Refer to the notes in tail_logs.
-        self.run_on_head(
-            handle,
-            code,
-            stream_logs=True,
-            process_stream=False,
-            ssh_mode=command_runner.SshMode.INTERACTIVE,
-        )
+        try:
+            returncode = self.run_on_head(
+                handle,
+                code,
+                stream_logs=True,
+                process_stream=False,
+                ssh_mode=command_runner.SshMode.INTERACTIVE,
+            )
+        except SystemExit as e:
+            returncode = e.code
+        return returncode
 
     def sync_down_managed_job_logs(
             self,

sky/cli.py

@@ -1227,11 +1227,15 @@ def launch(
             clusters=[handle.get_cluster_name()])
         # job_id will be None if no job was submitted (e.g. no entrypoint
         # provided)
+        returncode = 0
         if not detach_run and job_id is not None:
-            sdk.tail_logs(handle.get_cluster_name(), job_id, follow=True)
+            returncode = sdk.tail_logs(handle.get_cluster_name(),
+                                       job_id,
+                                       follow=True)
         click.secho(
             ux_utils.command_hint_messages(ux_utils.CommandHintType.CLUSTER_JOB,
                                            job_id, handle.get_cluster_name()))
+        sys.exit(returncode)
 
 
 @cli.command(cls=_DocumentedCodeCommand)
@@ -1377,7 +1381,8 @@ def exec(cluster: Optional[str], cluster_option: Optional[str],
     job_id_handle = _async_call_or_wait(request_id, async_call, 'sky.exec')
     if not async_call and not detach_run:
         job_id, _ = job_id_handle
-        sdk.tail_logs(cluster, job_id, follow=True)
+        returncode = sdk.tail_logs(cluster, job_id, follow=True)
+        sys.exit(returncode)
 
 
 def _handle_jobs_queue_request(
@@ -2121,12 +2126,20 @@ def logs(
     one job_id can be provided.
 
     2. If ``--status`` is specified, print the status of the job and exit with
-    returncode 0 if the job succeeded, or 1 otherwise. At most one job_id can
-    be specified.
+    returncode 0 if the job succeeded. At most one job_id can
+    be specified. Other possible return codes:
+
+    - 100: job failed.
+    - 101: job not finished.
+    - 102: job not found.
+    - 103: job was cancelled by the user.
 
     3. If ``--sync-down`` is specified, the logs of the job will be downloaded
     from the cluster and saved to the local machine under
-    ``~/sky_logs``. Mulitple job_ids can be specified.
+    ``~/sky_logs``. Multiple job_ids can be specified.
+
+    4. If the job fails or fetching the logs fails, the command will exit with
+    a non-zero return code.
     """
     if sync_down and status:
         raise click.UsageError(
@@ -2174,17 +2187,18 @@ def logs(
         # it will return {None: None}.
         if job_id is None:
             click.secho(f'No job found on cluster {cluster!r}.', fg='red')
-            sys.exit(1)
+            sys.exit(exceptions.JobExitCode.NOT_FOUND)
         job_status = list(job_statuses.values())[0]
         job_status_str = job_status.value if job_status is not None else 'None'
         click.echo(f'Job {job_id}: {job_status_str}')
         if job_status == job_lib.JobStatus.SUCCEEDED:
             return
         else:
+            returncode = exceptions.JobExitCode.from_job_status(job_status)
             if job_status is None:
                 id_str = '' if job_id is None else f'{job_id} '
                 click.secho(f'Job {id_str}not found', fg='red')
-            sys.exit(1)
+            sys.exit(returncode)
 
     job_str = f'job {job_id}'
     if job_id is None:
@@ -2194,7 +2208,8 @@ def logs(
                 f'{colorama.Style.RESET_ALL}')
 
     # Stream logs from the server.
-    sdk.tail_logs(cluster, job_id, follow, tail=tail)
+    returncode = sdk.tail_logs(cluster, job_id, follow, tail=tail)
+    sys.exit(returncode)
 
 
 @cli.command()
@@ -3019,7 +3034,7 @@ def _down_or_stop_clusters(
                     # with the termination.
                     hint_or_raise(controller_name, purge)
                 except (exceptions.ClusterOwnerIdentityMismatchError,
-                        RuntimeError) as e:
+                        exceptions.NotSupportedError, RuntimeError) as e:
                     if purge:
                         click.echo(common_utils.format_exception(e))
                     else:
@@ -3729,6 +3744,7 @@ def storage_delete(names: List[str], all: bool, yes: bool, async_call: bool):  #
         if not storages:
             click.echo('No storage(s) to delete.')
             return
+        names = [storage['name'] for storage in storages]
     else:
         names = _get_glob_storages(names)
     if names:
@@ -3893,10 +3909,11 @@ def jobs_launch(
                                         'sky.jobs.launch')
     if not async_call and not detach_run:
         job_id = job_id_handle[0]
-        managed_jobs.tail_logs(name=None,
-                               job_id=job_id,
-                               follow=True,
-                               controller=False)
+        returncode = managed_jobs.tail_logs(name=None,
+                                            job_id=job_id,
+                                            follow=True,
+                                            controller=False)
+        sys.exit(returncode)
 
 
 @jobs.command('queue', cls=_DocumentedCodeCommand)
@@ -4127,11 +4144,12 @@ def jobs_logs(name: Optional[str], job_id: Optional[int], follow: bool,
                 logger.info(f'{fore.CYAN}Job {job} logs{controller_str}: '
                             f'{log_local_path}{style.RESET_ALL}')
         else:
-            managed_jobs.tail_logs(name=name,
-                                   job_id=job_id,
-                                   follow=follow,
-                                   controller=controller,
-                                   refresh=refresh)
+            returncode = managed_jobs.tail_logs(name=name,
+                                                job_id=job_id,
+                                                follow=follow,
+                                                controller=controller,
+                                                refresh=refresh)
+            sys.exit(returncode)
     except exceptions.ClusterNotUpError:
         with ux_utils.print_exception_no_traceback():
             raise

sky/client/cli.py

@@ -1227,11 +1227,15 @@ def launch(
             clusters=[handle.get_cluster_name()])
         # job_id will be None if no job was submitted (e.g. no entrypoint
         # provided)
+        returncode = 0
         if not detach_run and job_id is not None:
-            sdk.tail_logs(handle.get_cluster_name(), job_id, follow=True)
+            returncode = sdk.tail_logs(handle.get_cluster_name(),
+                                       job_id,
+                                       follow=True)
         click.secho(
             ux_utils.command_hint_messages(ux_utils.CommandHintType.CLUSTER_JOB,
                                            job_id, handle.get_cluster_name()))
+        sys.exit(returncode)
 
 
 @cli.command(cls=_DocumentedCodeCommand)
@@ -1377,7 +1381,8 @@ def exec(cluster: Optional[str], cluster_option: Optional[str],
     job_id_handle = _async_call_or_wait(request_id, async_call, 'sky.exec')
     if not async_call and not detach_run:
         job_id, _ = job_id_handle
-        sdk.tail_logs(cluster, job_id, follow=True)
+        returncode = sdk.tail_logs(cluster, job_id, follow=True)
+        sys.exit(returncode)
 
 
 def _handle_jobs_queue_request(
@@ -2121,12 +2126,20 @@ def logs(
     one job_id can be provided.
 
     2. If ``--status`` is specified, print the status of the job and exit with
-    returncode 0 if the job succeeded, or 1 otherwise. At most one job_id can
-    be specified.
+    returncode 0 if the job succeeded. At most one job_id can
+    be specified. Other possible return codes:
+
+    - 100: job failed.
+    - 101: job not finished.
+    - 102: job not found.
+    - 103: job was cancelled by the user.
 
     3. If ``--sync-down`` is specified, the logs of the job will be downloaded
     from the cluster and saved to the local machine under
-    ``~/sky_logs``. Mulitple job_ids can be specified.
+    ``~/sky_logs``. Multiple job_ids can be specified.
+
+    4. If the job fails or fetching the logs fails, the command will exit with
+    a non-zero return code.
     """
     if sync_down and status:
         raise click.UsageError(
@@ -2174,17 +2187,18 @@ def logs(
         # it will return {None: None}.
         if job_id is None:
             click.secho(f'No job found on cluster {cluster!r}.', fg='red')
-            sys.exit(1)
+            sys.exit(exceptions.JobExitCode.NOT_FOUND)
         job_status = list(job_statuses.values())[0]
         job_status_str = job_status.value if job_status is not None else 'None'
         click.echo(f'Job {job_id}: {job_status_str}')
         if job_status == job_lib.JobStatus.SUCCEEDED:
             return
         else:
+            returncode = exceptions.JobExitCode.from_job_status(job_status)
             if job_status is None:
                 id_str = '' if job_id is None else f'{job_id} '
                 click.secho(f'Job {id_str}not found', fg='red')
-            sys.exit(1)
+            sys.exit(returncode)
 
     job_str = f'job {job_id}'
     if job_id is None:
@@ -2194,7 +2208,8 @@ def logs(
                 f'{colorama.Style.RESET_ALL}')
 
     # Stream logs from the server.
-    sdk.tail_logs(cluster, job_id, follow, tail=tail)
+    returncode = sdk.tail_logs(cluster, job_id, follow, tail=tail)
+    sys.exit(returncode)
 
 
 @cli.command()
@@ -3019,7 +3034,7 @@ def _down_or_stop_clusters(
                     # with the termination.
                     hint_or_raise(controller_name, purge)
                 except (exceptions.ClusterOwnerIdentityMismatchError,
-                        RuntimeError) as e:
+                        exceptions.NotSupportedError, RuntimeError) as e:
                     if purge:
                         click.echo(common_utils.format_exception(e))
                     else:
@@ -3729,6 +3744,7 @@ def storage_delete(names: List[str], all: bool, yes: bool, async_call: bool):  #
         if not storages:
             click.echo('No storage(s) to delete.')
             return
+        names = [storage['name'] for storage in storages]
     else:
         names = _get_glob_storages(names)
     if names:
@@ -3893,10 +3909,11 @@ def jobs_launch(
                                         'sky.jobs.launch')
     if not async_call and not detach_run:
         job_id = job_id_handle[0]
-        managed_jobs.tail_logs(name=None,
-                               job_id=job_id,
-                               follow=True,
-                               controller=False)
+        returncode = managed_jobs.tail_logs(name=None,
+                                            job_id=job_id,
+                                            follow=True,
+                                            controller=False)
+        sys.exit(returncode)
 
 
 @jobs.command('queue', cls=_DocumentedCodeCommand)
@@ -4127,11 +4144,12 @@ def jobs_logs(name: Optional[str], job_id: Optional[int], follow: bool,
                 logger.info(f'{fore.CYAN}Job {job} logs{controller_str}: '
                             f'{log_local_path}{style.RESET_ALL}')
         else:
-            managed_jobs.tail_logs(name=name,
-                                   job_id=job_id,
-                                   follow=follow,
-                                   controller=controller,
-                                   refresh=refresh)
+            returncode = managed_jobs.tail_logs(name=name,
+                                                job_id=job_id,
+                                                follow=follow,
+                                                controller=controller,
+                                                refresh=refresh)
+            sys.exit(returncode)
     except exceptions.ClusterNotUpError:
         with ux_utils.print_exception_no_traceback():
             raise

sky/client/sdk.py

@@ -25,6 +25,7 @@ import filelock
 import psutil
 import requests
 
+from sky import admin_policy
 from sky import backends
 from sky import exceptions
 from sky import sky_logging
@@ -212,13 +213,17 @@ def list_accelerator_counts(
 @annotations.client_api
 def optimize(
     dag: 'sky.Dag',
-    minimize: common.OptimizeTarget = common.OptimizeTarget.COST
+    minimize: common.OptimizeTarget = common.OptimizeTarget.COST,
+    admin_policy_request_options: Optional[admin_policy.RequestOptions] = None
 ) -> server_common.RequestId:
     """Finds the best execution plan for the given DAG.
 
     Args:
         dag: the DAG to optimize.
         minimize: whether to minimize cost or time.
+        admin_policy_request_options: Request options used for admin policy
+            validation. This is only required when a admin policy is in use,
+            see: https://docs.skypilot.co/en/latest/cloud-setup/policy.html
 
     Returns:
         The request ID of the optimize request.
@@ -233,7 +238,9 @@ def optimize(
     """
     dag_str = dag_utils.dump_chain_dag_to_yaml_str(dag)
 
-    body = payloads.OptimizeBody(dag=dag_str, minimize=minimize)
+    body = payloads.OptimizeBody(dag=dag_str,
+                                 minimize=minimize,
+                                 request_options=admin_policy_request_options)
     response = requests.post(f'{server_common.get_server_url()}/optimize',
                              json=json.loads(body.model_dump_json()))
     return server_common.get_request_id(response)
@@ -242,7 +249,11 @@ def optimize(
 @usage_lib.entrypoint
 @server_common.check_server_healthy_or_start
 @annotations.client_api
-def validate(dag: 'sky.Dag', workdir_only: bool = False) -> None:
+def validate(
+    dag: 'sky.Dag',
+    workdir_only: bool = False,
+    admin_policy_request_options: Optional[admin_policy.RequestOptions] = None
+) -> None:
     """Validates the tasks.
 
     The file paths (workdir and file_mounts) are validated on the client side
@@ -254,13 +265,17 @@ def validate(dag: 'sky.Dag', workdir_only: bool = False) -> None:
         dag: the DAG to validate.
         workdir_only: whether to only validate the workdir. This is used for
             `exec` as it does not need other files/folders in file_mounts.
+        admin_policy_request_options: Request options used for admin policy
+            validation. This is only required when a admin policy is in use,
+            see: https://docs.skypilot.co/en/latest/cloud-setup/policy.html
     """
     for task in dag.tasks:
         task.expand_and_validate_workdir()
         if not workdir_only:
             task.expand_and_validate_file_mounts()
     dag_str = dag_utils.dump_chain_dag_to_yaml_str(dag)
-    body = payloads.ValidateBody(dag=dag_str)
+    body = payloads.ValidateBody(dag=dag_str,
+                                 request_options=admin_policy_request_options)
     response = requests.post(f'{server_common.get_server_url()}/validate',
                              json=json.loads(body.model_dump_json()))
     if response.status_code == 400:
@@ -386,7 +401,12 @@ def launch(
                                       'Please contact the SkyPilot team if you '
                                       'need this feature at slack.skypilot.co.')
     dag = dag_utils.convert_entrypoint_to_dag(task)
-    validate(dag)
+    request_options = admin_policy.RequestOptions(
+        cluster_name=cluster_name,
+        idle_minutes_to_autostop=idle_minutes_to_autostop,
+        down=down,
+        dryrun=dryrun)
+    validate(dag, admin_policy_request_options=request_options)
 
     confirm_shown = False
     if _need_confirmation:
@@ -400,7 +420,8 @@ def launch(
         if not clusters:
             # Show the optimize log before the prompt if the cluster does not
             # exist.
-            request_id = optimize(dag)
+            request_id = optimize(dag,
+                                  admin_policy_request_options=request_options)
             stream_and_get(request_id)
         else:
             cluster_record = clusters[0]
@@ -562,7 +583,7 @@ def tail_logs(cluster_name: str,
               job_id: Optional[int],
               follow: bool,
               tail: int = 0,
-              output_stream: Optional['io.TextIOBase'] = None) -> None:
+              output_stream: Optional['io.TextIOBase'] = None) -> int:
     """Tails the logs of a job.
 
     Args:
@@ -575,7 +596,9 @@ def tail_logs(cluster_name: str,
             console.
 
     Returns:
-        None
+        Exit code based on success or failure of the job. 0 if success,
+        100 if the job failed. See exceptions.JobExitCode for possible exit
+        codes.
 
     Request Raises:
         ValueError: if arguments are invalid or the cluster is not supported.
@@ -601,7 +624,7 @@ def tail_logs(cluster_name: str,
         timeout=(client_common.API_SERVER_REQUEST_CONNECTION_TIMEOUT_SECONDS,
                  None))
     request_id = server_common.get_request_id(response)
-    stream_response(request_id, response, output_stream)
+    return stream_response(request_id, response, output_stream)
 
 
 @usage_lib.entrypoint

sky/core.py

@@ -6,16 +6,18 @@ from typing import Any, Dict, List, Optional, Tuple, Union
 
 import colorama
 
+from sky import admin_policy
 from sky import backends
 from sky import check as sky_check
 from sky import clouds
-from sky import dag
+from sky import dag as dag_lib
 from sky import data
 from sky import exceptions
 from sky import global_user_state
 from sky import models
+from sky import optimizer
 from sky import sky_logging
-from sky import task
+from sky import task as task_lib
 from sky.backends import backend_utils
 from sky.clouds import service_catalog
 from sky.jobs.server import core as managed_jobs_core
@@ -25,6 +27,7 @@ from sky.skylet import constants
 from sky.skylet import job_lib
 from sky.skylet import log_lib
 from sky.usage import usage_lib
+from sky.utils import admin_policy_utils
 from sky.utils import common
 from sky.utils import common_utils
 from sky.utils import controller_utils
@@ -44,6 +47,46 @@ logger = sky_logging.init_logger(__name__)
 # ======================
 
 
+@usage_lib.entrypoint
+def optimize(
+    dag: 'dag_lib.Dag',
+    minimize: common.OptimizeTarget = common.OptimizeTarget.COST,
+    blocked_resources: Optional[List['resources_lib.Resources']] = None,
+    quiet: bool = False,
+    request_options: Optional[admin_policy.RequestOptions] = None
+) -> 'dag_lib.Dag':
+    """Finds the best execution plan for the given DAG.
+
+    Args:
+        dag: the DAG to optimize.
+        minimize: whether to minimize cost or time.
+        blocked_resources: a list of resources that should not be used.
+        quiet: whether to suppress logging.
+        request_options: Request options used in enforcing admin policies.
+            This is only required when a admin policy is in use,
+            see: https://docs.skypilot.co/en/latest/cloud-setup/policy.html
+    Returns:
+        The optimized DAG.
+
+    Raises:
+        exceptions.ResourcesUnavailableError: if no resources are available
+            for a task.
+        exceptions.NoCloudAccessError: if no public clouds are enabled.
+    """
+    # TODO: We apply the admin policy only on the first DAG optimization which
+    # is shown on `sky launch`. The optimizer is also invoked during failover,
+    # but we do not apply the admin policy there. We should apply the admin
+    # policy in the optimizer, but that will require some refactoring.
+    dag, _ = admin_policy_utils.apply(
+        dag,
+        use_mutated_config_in_current_request=True,
+        request_options=request_options)
+    return optimizer.Optimizer.optimize(dag=dag,
+                                        minimize=minimize,
+                                        blocked_resources=blocked_resources,
+                                        quiet=quiet)
+
+
 @usage_lib.entrypoint
 def status(
     cluster_names: Optional[Union[str, List[str]]] = None,
@@ -325,8 +368,8 @@ def _start(
 
     usage_lib.record_cluster_name_for_current_operation(cluster_name)
 
-    with dag.Dag():
-        dummy_task = task.Task().set_resources(handle.launched_resources)
+    with dag_lib.Dag():
+        dummy_task = task_lib.Task().set_resources(handle.launched_resources)
         dummy_task.num_nodes = handle.launched_nodes
     handle = backend.provision(dummy_task,
                                to_provision=handle.launched_resources,
@@ -783,7 +826,7 @@ def cancel(
 def tail_logs(cluster_name: str,
               job_id: Optional[int],
               follow: bool = True,
-              tail: int = 0) -> None:
+              tail: int = 0) -> int:
     # NOTE(dev): Keep the docstring consistent between the Python API and CLI.
     """Tails the logs of a job.
 
@@ -799,6 +842,12 @@ def tail_logs(cluster_name: str,
           not the same as the user who created the cluster.
         sky.exceptions.CloudUserIdentityError: if we fail to get the current
           user identity.
+
+    Returns:
+        Return code based on success or failure of the job. 0 if success,
+          100 if the job failed. Note: This is not the return code of the job
+          script.
+
     """
     # Check the status of the cluster.
     handle = backend_utils.check_cluster_available(
@@ -808,7 +857,7 @@ def tail_logs(cluster_name: str,
     backend = backend_utils.get_backend_from_handle(handle)
 
     usage_lib.record_cluster_name_for_current_operation(cluster_name)
-    backend.tail_logs(handle, job_id, follow=follow, tail=tail)
+    return backend.tail_logs(handle, job_id, follow=follow, tail=tail)
 
 
 @usage_lib.entrypoint

sky/exceptions.py

@@ -9,7 +9,9 @@ from typing import Any, Dict, List, Optional, Sequence
 from sky.utils import env_options
 
 if typing.TYPE_CHECKING:
+    from sky import jobs as managed_jobs
     from sky.backends import backend
+    from sky.skylet import job_lib
     from sky.utils import status_lib
 
 # Return code for keyboard interruption and SIGTSTP
@@ -236,7 +238,7 @@ class CommandError(SkyPilotExcludeArgsBaseException):
         else:
             if (len(command) > 100 and
                     not env_options.Options.SHOW_DEBUG_INFO.get()):
-                # Chunck the command to avoid overflow.
+                # Chunk the command to avoid overflow.
                 command = command[:100] + '...'
             message = (f'Command {command} failed with return code '
                        f'{returncode}.\n{error_msg}')
@@ -449,3 +451,80 @@ class ApiServerConnectionError(RuntimeError):
             f'Could not connect to SkyPilot API server at {server_url}. '
             f'Please ensure that the server is running. '
             f'Try: curl {server_url}/api/health')
+
+
+class JobExitCode(enum.IntEnum):
+    """Job exit code enum.
+
+    These codes are used as return codes for job-related operations and as
+    process exit codes to indicate job status.
+    """
+
+    SUCCEEDED = 0
+    """The job completed successfully"""
+
+    FAILED = 100
+    """The job failed (due to user code, setup, or driver failure)"""
+
+    NOT_FINISHED = 101
+    """The job has not finished yet"""
+
+    NOT_FOUND = 102
+    """The job was not found"""
+
+    CANCELLED = 103
+    """The job was cancelled by the user"""
+
+    @classmethod
+    def from_job_status(cls,
+                        status: Optional['job_lib.JobStatus']) -> 'JobExitCode':
+        """Convert a job status to an exit code."""
+        # Import here to avoid circular imports
+        # pylint: disable=import-outside-toplevel
+        from sky.skylet import job_lib
+
+        if status is None:
+            return cls.NOT_FOUND
+
+        if not status.is_terminal():
+            return cls.NOT_FINISHED
+
+        if status == job_lib.JobStatus.SUCCEEDED:
+            return cls.SUCCEEDED
+
+        if status == job_lib.JobStatus.CANCELLED:
+            return cls.CANCELLED
+
+        if status in job_lib.JobStatus.user_code_failure_states(
+        ) or status == job_lib.JobStatus.FAILED_DRIVER:
+            return cls.FAILED
+
+        # Should not hit this case, but included to avoid errors
+        return cls.FAILED
+
+    @classmethod
+    def from_managed_job_status(
+            cls,
+            status: Optional['managed_jobs.ManagedJobStatus']) -> 'JobExitCode':
+        """Convert a managed job status to an exit code."""
+        # Import here to avoid circular imports
+        # pylint: disable=import-outside-toplevel
+        from sky import jobs as managed_jobs
+
+        if status is None:
+            return cls.NOT_FOUND
+
+        if not status.is_terminal():
+            return cls.NOT_FINISHED
+
+        if status == managed_jobs.ManagedJobStatus.SUCCEEDED:
+            return cls.SUCCEEDED
+
+        if status == managed_jobs.ManagedJobStatus.CANCELLED:
+            return cls.CANCELLED
+
+        if status.is_failed():
+            return cls.FAILED
+
+        # Should not hit this case, but included to avoid errors
+        return cls.FAILED