o civ@sddlZddlZddlmZmZmZddlmZmZddl m Z ddl m Z e eZ  dde deded eed eed ef d d Ze dddedededee dededeeefd efddZeZdS)N)AnyDictOptional)CONTROL_LOOP_INTERVAL_SSERVE_LOGGER_NAME)AutoscalingConfig) PublicAPIautoscaling_configtotal_num_requestsnum_running_replicasoverride_min_replicasoverride_max_replicasreturnc Cs|dkrtd||}||}|dkr|}n|}d|d|}t||} t|||kr?| |kr?| d8} |j} |j} |durK|} |durQ|} t| t | | } | S)a8Returns the number of replicas to scale to based on the given metrics. Args: autoscaling_config: The autoscaling parameters to use for this calculation. current_num_ongoing_requests (List[float]): A list of the number of ongoing requests for each replica. Assumes each entry has already been time-averaged over the desired lookback window. override_min_replicas: Overrides min_replicas from the config when calculating the final number of replicas. override_max_replicas: Overrides max_replicas from the config when calculating the final number of replicas. Returns: desired_num_replicas: The desired number of replicas to scale to, based on the input metrics and the current number of replicas. rz!Number of replicas cannot be zeroN) ValueErrorget_target_ongoing_requestsget_upscaling_factorget_downscaling_factormathceil min_replicas max_replicasmaxmin) r r r r r target_num_requests error_ratioscaling_factorsmoothed_error_ratiodesired_num_replicasrrrP/home/ubuntu/.local/lib/python3.10/site-packages/ray/serve/autoscaling_policy.py_calculate_desired_num_replicas s*  r!alpha) stabilitycurr_target_num_replicasconfigcapacity_adjusted_min_replicascapacity_adjusted_max_replicas policy_statec Cs|dd}|dkr|dkrttd||S|S|}t|||||d} | |krC|dkr1d}|d7}|t|jtkrBd}| }n| |kr`|dkrMd}|d8}|t|j t kr_d}| }nd}||d<|S)aThe default autoscaling policy based on basic thresholds for scaling. There is a minimum threshold for the average queue length in the cluster to scale up and a maximum threshold to scale down. Each period, a 'scale up' or 'scale down' decision is made. This decision must be made for a specified number of periods in a row before the number of replicas is actually scaled. See config options for more details. Assumes `get_decision_num_replicas` is called once every CONTROL_LOOP_PERIOD_S seconds. decision_counterrr)r r r ) getrrrrr!intupscale_delay_srdownscale_delay_s) r$r r r%r&r'r(r)decision_num_replicasrrrr 'replica_queue_length_autoscaling_policyTsF r/)NN)loggingrtypingrrrray.serve._private.constantsrrray.serve.configrray.util.annotationsr getLoggerloggerr+floatr!strr/default_autoscaling_policyrrrr sR    H  J