skypilot-nightly 1.0.0.dev20251009__py3-none-any.whl → 1.0.0.dev20251107__py3-none-any.whl

sky/server/requests/preconditions.py

@@ -90,7 +90,7 @@ class Precondition(abc.ABC):
         while True:
             if self.timeout > 0 and time.time() - start_time > self.timeout:
                 # Cancel the request on timeout.
-                api_requests.set_request_failed(
+                await api_requests.set_request_failed_async(
                     self.request_id,
                     exceptions.RequestCancelled(
                         f'Request {self.request_id} precondition wait timed '
@@ -98,13 +98,15 @@ class Precondition(abc.ABC):
                 return False
 
             # Check if the request has been cancelled
-            request = await api_requests.get_request_async(self.request_id)
+            request = await api_requests.get_request_async(self.request_id,
+                                                           fields=['status'])
             if request is None:
                 logger.error(f'Request {self.request_id} not found')
                 return False
             if request.status == api_requests.RequestStatus.CANCELLED:
                 logger.debug(f'Request {self.request_id} cancelled')
                 return False
+            del request
 
             try:
                 met, status_msg = await self.check()
@@ -112,13 +114,11 @@ class Precondition(abc.ABC):
                     return True
                 if status_msg is not None and status_msg != last_status_msg:
                     # Update the status message if it has changed.
-                    async with api_requests.update_request_async(
-                            self.request_id) as req:
-                        assert req is not None, self.request_id
-                        req.status_msg = status_msg
+                    await api_requests.update_status_msg_async(
+                        self.request_id, status_msg)
                     last_status_msg = status_msg
             except (Exception, SystemExit, KeyboardInterrupt) as e:  # pylint: disable=broad-except
-                api_requests.set_request_failed(self.request_id, e)
+                await api_requests.set_request_failed_async(self.request_id, e)
                 logger.info(f'Request {self.request_id} failed due to '
                             f'{common_utils.format_exception(e)}')
                 return False
@@ -164,11 +164,14 @@ class ClusterStartCompletePrecondition(Precondition):
         requests = await api_requests.get_request_tasks_async(
             req_filter=api_requests.RequestTaskFilter(
                 status=[
-                    api_requests.RequestStatus.RUNNING,
-                    api_requests.RequestStatus.PENDING
+                    api_requests.RequestStatus.PENDING,
+                    api_requests.RequestStatus.RUNNING
                 ],
                 include_request_names=['sky.launch', 'sky.start'],
-                cluster_names=[self.cluster_name]))
+                cluster_names=[self.cluster_name],
+                # Only get the request ID to avoid fetching the whole request.
+                # We're only interested in the count, not the whole request.
+                fields=['request_id']))
         if len(requests) == 0:
             # No running or pending tasks, the start process is done.
             return True, None

sky/server/requests/request_names.py

@@ -0,0 +1,120 @@
+"""Request names."""
+import enum
+
+
+class RequestName(str, enum.Enum):
+    """Enum of all the request names."""
+    # General requests
+    CHECK = 'check'
+    ENABLED_CLOUDS = 'enabled_clouds'
+    REALTIME_KUBERNETES_GPU_AVAILABILITY = (
+        'realtime_kubernetes_gpu_availability')
+    KUBERNETES_NODE_INFO = 'kubernetes_node_info'
+    STATUS_KUBERNETES = 'status_kubernetes'
+    LIST_ACCELERATORS = 'list_accelerators'
+    LIST_ACCELERATOR_COUNTS = 'list_accelerator_counts'
+    OPTIMIZE = 'optimize'
+    # Cluster requests
+    CLUSTER_LAUNCH = 'launch'
+    CLUSTER_EXEC = 'exec'
+    CLUSTER_STOP = 'stop'
+    CLUSTER_STATUS = 'status'
+    CLUSTER_ENDPOINTS = 'endpoints'
+    CLUSTER_DOWN = 'down'
+    CLUSTER_START = 'start'
+    CLUSTER_AUTOSTOP = 'autostop'
+    CLUSTER_QUEUE = 'queue'
+    CLUSTER_JOB_STATUS = 'job_status'
+    CLUSTER_JOB_CANCEL = 'cancel'
+    CLUSTER_JOB_LOGS = 'logs'
+    CLUSTER_JOB_DOWNLOAD_LOGS = 'download_logs'
+    CLUSTER_COST_REPORT = 'cost_report'
+    # Storage requests
+    STORAGE_LS = 'storage_ls'
+    STORAGE_DELETE = 'storage_delete'
+    # Local requests
+    LOCAL_UP = 'local_up'
+    LOCAL_DOWN = 'local_down'
+    # API requests
+    API_CANCEL = 'api_cancel'
+    ALL_CONTEXTS = 'all_contexts'
+    # Managed jobs requests
+    JOBS_LAUNCH = 'jobs.launch'
+    JOBS_QUEUE = 'jobs.queue'
+    JOBS_QUEUE_V2 = 'jobs.queue_v2'
+    JOBS_CANCEL = 'jobs.cancel'
+    JOBS_LOGS = 'jobs.logs'
+    JOBS_DOWNLOAD_LOGS = 'jobs.download_logs'
+    JOBS_POOL_APPLY = 'jobs.pool_apply'
+    JOBS_POOL_DOWN = 'jobs.pool_down'
+    JOBS_POOL_STATUS = 'jobs.pool_status'
+    JOBS_POOL_LOGS = 'jobs.pool_logs'
+    JOBS_POOL_SYNC_DOWN_LOGS = 'jobs.pool_sync_down_logs'
+    # Serve requests
+    SERVE_UP = 'serve.up'
+    SERVE_UPDATE = 'serve.update'
+    SERVE_DOWN = 'serve.down'
+    SERVE_TERMINATE_REPLICA = 'serve.terminate_replica'
+    SERVE_STATUS = 'serve.status'
+    SERVE_LOGS = 'serve.logs'
+    SERVE_SYNC_DOWN_LOGS = 'serve.sync_down_logs'
+    # Volumes requests
+    VOLUME_LIST = 'volume_list'
+    VOLUME_DELETE = 'volume_delete'
+    VOLUME_APPLY = 'volume_apply'
+    # Workspaces requests
+    WORKSPACES_GET = 'workspaces.get'
+    WORKSPACES_UPDATE = 'workspaces.update'
+    WORKSPACES_CREATE = 'workspaces.create'
+    WORKSPACES_DELETE = 'workspaces.delete'
+    WORKSPACES_GET_CONFIG = 'workspaces.get_config'
+    WORKSPACES_UPDATE_CONFIG = 'workspaces.update_config'
+    # SSH node pools requests
+    SSH_NODE_POOLS_UP = 'ssh_node_pools.up'
+    SSH_NODE_POOLS_DOWN = 'ssh_node_pools.down'
+    # Internal request daemons
+    REQUEST_DAEMON_STATUS_REFRESH = 'status-refresh'
+    REQUEST_DAEMON_VOLUME_REFRESH = 'volume-refresh'
+    REQUEST_DAEMON_MANAGED_JOB_STATUS_REFRESH = 'managed-job-status-refresh'
+    REQUEST_DAEMON_SKY_SERVE_STATUS_REFRESH = 'sky-serve-status-refresh'
+    REQUEST_DAEMON_POOL_STATUS_REFRESH = 'pool-status-refresh'
+
+    def __repr__(self):
+        return self.value
+
+    def __str__(self):
+        return self.value
+
+
+class AdminPolicyRequestName(str, enum.Enum):
+    """Enum of all the request names that are
+    used for admin policy application."""
+    # General requests
+    # validate call is not stored in the request db,
+    # but is defined here for admin policy application
+    VALIDATE = 'validate'
+    OPTIMIZE = RequestName.OPTIMIZE.value
+    # Cluster requests
+    CLUSTER_LAUNCH = RequestName.CLUSTER_LAUNCH.value
+    CLUSTER_EXEC = RequestName.CLUSTER_EXEC.value
+    # Jobs requests
+    JOBS_LAUNCH = RequestName.JOBS_LAUNCH.value
+    # jobs launch controller request is not stored in the request db,
+    # but is defined here for admin policy application
+    JOBS_LAUNCH_CONTROLLER = 'jobs.launch_controller'
+    JOBS_POOL_APPLY = RequestName.JOBS_POOL_APPLY.value
+    # Serve requests
+    SERVE_UP = RequestName.SERVE_UP.value
+    # serve launch controller request is not stored in the request db,
+    # but is defined here for admin policy application
+    SERVE_LAUNCH_CONTROLLER = 'serve.launch_controller'
+    SERVE_UPDATE = RequestName.SERVE_UPDATE.value
+    # serve launch replica request is not stored in the request db,
+    # but is defined here for admin policy application
+    SERVE_LAUNCH_REPLICA = 'serve.launch_replica'
+
+    def __repr__(self):
+        return self.value
+
+    def __str__(self):
+        return self.value