def __init__( self, controller_handle: ActorHandle, deployment_name: EndpointTag, handle_options: Optional[HandleOptions] = None, *, _router: Optional[Router] = None, _internal_pickled_http_request: bool = False, ): self.controller_handle = controller_handle self.deployment_name = deployment_name self.handle_options = handle_options or HandleOptions() self.handle_tag = f"{self.deployment_name}#{get_random_letters()}" self._pickled_http_request = _internal_pickled_http_request self.request_counter = metrics.Counter( "serve_handle_request_counter", description=("The number of handle.remote() calls that have been " "made on this handle."), tag_keys=("handle", "deployment"), ) self.request_counter.set_default_tags({ "handle": self.handle_tag, "deployment": self.deployment_name }) self.router: Router = _router or self._make_router() deployment_route = DeploymentRoute.FromString( ray.get( self.controller_handle.get_deployment_info.remote( self.deployment_name))) deployment_info = DeploymentInfo.from_proto( deployment_route.deployment_info) self._stop_event: Optional[threading.Event] = None self._pusher: Optional[threading.Thread] = None remote_func = self.controller_handle.record_handle_metrics.remote if deployment_info.deployment_config.autoscaling_config: self._stop_event = threading.Event() self._pusher = start_metrics_pusher( interval_s=HANDLE_METRIC_PUSH_INTERVAL_S, collection_callback=self._collect_handle_queue_metrics, metrics_process_func=remote_func, stop_event=self._stop_event, )
def __init__( self, _callable: Callable, deployment_name: str, replica_tag: ReplicaTag, deployment_config: DeploymentConfig, user_config: Any, version: DeploymentVersion, is_function: bool, controller_handle: ActorHandle, ) -> None: self.deployment_config = deployment_config self.deployment_name = deployment_name self.replica_tag = replica_tag self.callable = _callable self.is_function = is_function self.user_config = user_config self.version = version self.rwlock = aiorwlock.RWLock() user_health_check = getattr(_callable, HEALTH_CHECK_METHOD, None) if not callable(user_health_check): def user_health_check(): pass self.user_health_check = sync_to_async(user_health_check) self.num_ongoing_requests = 0 self.request_counter = metrics.Counter( "serve_deployment_request_counter", description= ("The number of queries that have been processed in this replica." ), tag_keys=("deployment", "replica"), ) self.request_counter.set_default_tags({ "deployment": self.deployment_name, "replica": self.replica_tag }) self.error_counter = metrics.Counter( "serve_deployment_error_counter", description=( "The number of exceptions that have occurred in this replica." ), tag_keys=("deployment", "replica"), ) self.error_counter.set_default_tags({ "deployment": self.deployment_name, "replica": self.replica_tag }) self.restart_counter = metrics.Counter( "serve_deployment_replica_starts", description= ("The number of times this replica has been restarted due to failure." ), tag_keys=("deployment", "replica"), ) self.restart_counter.set_default_tags({ "deployment": self.deployment_name, "replica": self.replica_tag }) self.processing_latency_tracker = metrics.Histogram( "serve_deployment_processing_latency_ms", description="The latency for queries to be processed.", boundaries=DEFAULT_LATENCY_BUCKET_MS, tag_keys=("deployment", "replica"), ) self.processing_latency_tracker.set_default_tags({ "deployment": self.deployment_name, "replica": self.replica_tag }) self.num_processing_items = metrics.Gauge( "serve_replica_processing_queries", description="The current number of queries being processed.", tag_keys=("deployment", "replica"), ) self.num_processing_items.set_default_tags({ "deployment": self.deployment_name, "replica": self.replica_tag }) self.restart_counter.inc() self._shutdown_wait_loop_s = deployment_config.graceful_shutdown_wait_loop_s if deployment_config.autoscaling_config: process_remote_func = controller_handle.record_autoscaling_metrics.remote config = deployment_config.autoscaling_config start_metrics_pusher( interval_s=config.metrics_interval_s, collection_callback=self._collect_autoscaling_metrics, metrics_process_func=process_remote_func, ) # NOTE(edoakes): we used to recommend that users use the "ray" logger # and tagged the logs with metadata as below. We now recommend using # the "ray.serve" 'component logger' (as of Ray 1.13). This is left to # maintain backwards compatibility with users who were using the # existing logger. We can consider removing it in Ray 2.0. ray_logger = logging.getLogger("ray") for handler in ray_logger.handlers: handler.setFormatter( logging.Formatter( handler.formatter._fmt + f" component=serve deployment={self.deployment_name} " f"replica={self.replica_tag}"))
def __init__( self, _callable: Callable, deployment_name: str, replica_tag: ReplicaTag, deployment_config: DeploymentConfig, user_config: Any, version: DeploymentVersion, is_function: bool, controller_handle: ActorHandle, ) -> None: self.deployment_config = deployment_config self.deployment_name = deployment_name self.replica_tag = replica_tag self.callable = _callable self.is_function = is_function self.user_config = user_config self.version = version self.rwlock = aiorwlock.RWLock() user_health_check = getattr(_callable, HEALTH_CHECK_METHOD, None) if not callable(user_health_check): def user_health_check(): pass self.user_health_check = sync_to_async(user_health_check) self.num_ongoing_requests = 0 self.request_counter = metrics.Counter( "serve_deployment_request_counter", description=("The number of queries that have been " "processed in this replica."), tag_keys=("deployment", "replica"), ) self.request_counter.set_default_tags({ "deployment": self.deployment_name, "replica": self.replica_tag }) self.error_counter = metrics.Counter( "serve_deployment_error_counter", description=("The number of exceptions that have " "occurred in this replica."), tag_keys=("deployment", "replica"), ) self.error_counter.set_default_tags({ "deployment": self.deployment_name, "replica": self.replica_tag }) self.restart_counter = metrics.Counter( "serve_deployment_replica_starts", description=("The number of times this replica " "has been restarted due to failure."), tag_keys=("deployment", "replica"), ) self.restart_counter.set_default_tags({ "deployment": self.deployment_name, "replica": self.replica_tag }) self.processing_latency_tracker = metrics.Histogram( "serve_deployment_processing_latency_ms", description="The latency for queries to be processed.", boundaries=DEFAULT_LATENCY_BUCKET_MS, tag_keys=("deployment", "replica"), ) self.processing_latency_tracker.set_default_tags({ "deployment": self.deployment_name, "replica": self.replica_tag }) self.num_processing_items = metrics.Gauge( "serve_replica_processing_queries", description="The current number of queries being processed.", tag_keys=("deployment", "replica"), ) self.num_processing_items.set_default_tags({ "deployment": self.deployment_name, "replica": self.replica_tag }) self.restart_counter.inc() self._shutdown_wait_loop_s = deployment_config.graceful_shutdown_wait_loop_s if deployment_config.autoscaling_config: config = deployment_config.autoscaling_config start_metrics_pusher( interval_s=config.metrics_interval_s, collection_callback=self._collect_autoscaling_metrics, controller_handle=controller_handle, ) ray_logger = logging.getLogger("ray") for handler in ray_logger.handlers: handler.setFormatter( logging.Formatter( handler.formatter._fmt + f" component=serve deployment={self.deployment_name} " f"replica={self.replica_tag}"))
def __init__(self, _callable: Callable, backend_tag: BackendTag, replica_tag: ReplicaTag, backend_config: BackendConfig, user_config: Any, version: BackendVersion, is_function: bool, controller_handle: ActorHandle) -> None: self.backend_tag = backend_tag self.replica_tag = replica_tag self.callable = _callable self.is_function = is_function self.backend_config = backend_config self.user_config = user_config self.version = version self.num_ongoing_requests = 0 self.request_counter = metrics.Counter( "serve_deployment_request_counter", description=("The number of queries that have been " "processed in this replica."), tag_keys=("deployment", "replica")) self.request_counter.set_default_tags({ "deployment": self.backend_tag, "replica": self.replica_tag }) self.loop = asyncio.get_event_loop() self.long_poll_client = LongPollClient( controller_handle, { (LongPollNamespace.BACKEND_CONFIGS, self.backend_tag): self._update_backend_configs, }, call_in_event_loop=self.loop, ) self.error_counter = metrics.Counter( "serve_deployment_error_counter", description=("The number of exceptions that have " "occurred in this replica."), tag_keys=("deployment", "replica")) self.error_counter.set_default_tags({ "deployment": self.backend_tag, "replica": self.replica_tag }) self.restart_counter = metrics.Counter( "serve_deployment_replica_starts", description=("The number of times this replica " "has been restarted due to failure."), tag_keys=("deployment", "replica")) self.restart_counter.set_default_tags({ "deployment": self.backend_tag, "replica": self.replica_tag }) self.processing_latency_tracker = metrics.Histogram( "serve_deployment_processing_latency_ms", description="The latency for queries to be processed.", boundaries=DEFAULT_LATENCY_BUCKET_MS, tag_keys=("deployment", "replica")) self.processing_latency_tracker.set_default_tags({ "deployment": self.backend_tag, "replica": self.replica_tag }) self.num_processing_items = metrics.Gauge( "serve_replica_processing_queries", description="The current number of queries being processed.", tag_keys=("deployment", "replica")) self.num_processing_items.set_default_tags({ "deployment": self.backend_tag, "replica": self.replica_tag }) self.restart_counter.inc() if backend_config.autoscaling_config: config = backend_config.autoscaling_config start_metrics_pusher( interval_s=config.metrics_interval_s, collection_callback=self._collect_autoscaling_metrics, controller_handle=controller_handle) ray_logger = logging.getLogger("ray") for handler in ray_logger.handlers: handler.setFormatter( logging.Formatter( handler.formatter._fmt + f" component=serve deployment={self.backend_tag} " f"replica={self.replica_tag}"))