timing.tracker
==============

.. py:module:: timing.tracker


Attributes
----------

.. autoapisummary::

   timing.tracker.AnomalyCallback


Classes
-------

.. autoapisummary::

   timing.tracker.StepTimingTracker


Module Contents
---------------

.. py:data:: AnomalyCallback

.. py:class:: StepTimingTracker(config: Optional[timing.config.TimingConfig] = None)

   Bases: :py:obj:`raccoon.class_name_logger.ClassNameLogger`


   Singleton tracker that persists step runtimes and reports anomalies.


   .. py:attribute:: config


   .. py:attribute:: database


   .. py:attribute:: anomaly_callbacks
      :type:  List[AnomalyCallback]
      :value: []



   .. py:method:: get_instance() -> StepTimingTracker
      :classmethod:


      Return the process-wide tracker instance used by ``Step.run_step``.



   .. py:method:: configure(config: timing.config.TimingConfig) -> None

      Replace configuration at runtime.



   .. py:method:: register_anomaly_callback(callback: AnomalyCallback) -> None

      Register an async callback to run when an anomaly is detected.



   .. py:method:: record_execution(signature: str, duration: float) -> Optional[timing.models.AnomalyDetection]
      :async:


      Persist execution and evaluate for anomalies.

      :returns: The detected anomaly, or ``None`` if the execution was normal.



   .. py:method:: get_upper_bound(signature: str) -> Optional[float]
      :async:


      Return the anomaly upper bound for a signature, or ``None`` if no baseline.

      This is the duration above which an execution would be flagged as
      SLOWER than expected.  Used by the live watchdog in ``Step.run_step``.