feature: ability to pass sampling_rate to logger setup so calls may get sampled in "debug" setup.

Author: danilohgds

.gitignore

@@ -291,5 +291,5 @@ $RECYCLE.BIN/
 # End of https://www.gitignore.io/api/osx,linux,python,windows,pycharm,visualstudiocode
 
 # Misc
-
-test_report
+test_report
+/.idea/*

python/aws_lambda_powertools/logging/logger.py

@@ -2,19 +2,22 @@
 import itertools
 import logging
 import os
+import random
 from distutils.util import strtobool
 from typing import Any, Callable, Dict
 
-from . import aws_lambda_logging
 from ..helper.models import MetricUnit, build_lambda_context_model, build_metric_unit_from_str
+from . import aws_lambda_logging
 
 logger = logging.getLogger(__name__)
 logger.setLevel(os.getenv("LOG_LEVEL", "INFO"))
 
 is_cold_start = True
 
 
-def logger_setup(service: str = "service_undefined", level: str = "INFO", **kwargs):
+def logger_setup(
+ service: str = "service_undefined", level: str = "INFO", sampling_rate: float = 0.0, **kwargs
+):
  """Setups root logger to format statements in JSON.
 
  Includes service name and any additional key=value into logs
@@ -26,13 +29,17 @@ def logger_setup(service: str = "service_undefined", level: str = "INFO", **kwar
  service name
  LOG_LEVEL: str
  logging level (e.g. INFO, DEBUG)
+ POWERTOOLS_SAMPLE_RATE: str
+ samping rate ranging from 0 to 1, float precision
 
  Parameters
  ----------
  service : str, optional
  service name to be appended in logs, by default "service_undefined"
  level : str, optional
  logging.level, by default "INFO"
+ sample_rate: float, optional
+ sample rate for debug calls within execution context defaults to 0
 
  Example
  -------
@@ -52,21 +59,36 @@ def logger_setup(service: str = "service_undefined", level: str = "INFO", **kwar
  >>>
  >>> def handler(event, context):
  logger.info("Hello")
+ :param service:
+ :param level:
+ :param sampling_rate:
 
  """
  service = os.getenv("POWERTOOLS_SERVICE_NAME") or service
+ sampling_rate = os.getenv("POWERTOOLS_SAMPLE_RATE") or sampling_rate
  log_level = os.getenv("LOG_LEVEL") or level
  logger = logging.getLogger(name=service)
+
+ # sampling a small percentage of requests with debug level, using a float value 0.1 = 10%~
+
+ try:
+ if sampling_rate and random.random() <= float(sampling_rate):
+ log_level = logging.DEBUG
+ except ValueError:
+ logger.debug("POWERTOOLS_SAMPLE_RATE provided value {0} is not valid.".format(sampling_rate))
+
  logger.setLevel(log_level)
 
  # Patch logger by structuring its outputs as JSON
- aws_lambda_logging.setup(level=log_level, service=service, **kwargs)
+ aws_lambda_logging.setup(
+ level=log_level, service=service, sampling_rate=sampling_rate, **kwargs
+ )
 
  return logger
 
 
 def logger_inject_lambda_context(
- lambda_handler: Callable[[Dict, Any], Any] = None, log_event: bool = False
+  lambda_handler: Callable[[Dict, Any], Any] = None, log_event: bool = False
 ):
  """Decorator to capture Lambda contextual info and inject into struct logging
 
@@ -110,6 +132,8 @@ def logger_inject_lambda_context(
  -------
  decorate : Callable
  Decorated lambda handler
+ :param log_event:
+ :param lambda_handler:
  """
 
  # If handler is None we've been called with parameters
@@ -159,12 +183,12 @@ def __is_cold_start() -> str:
 
 
 def log_metric(
- name: str,
- namespace: str,
- unit: MetricUnit,
- value: float = 0,
- service: str = "service_undefined",
- **dimensions,
+  name: str,
+  namespace: str,
+  unit: MetricUnit,
+  value: float = 0,
+  service: str = "service_undefined",
+  **dimensions,
 ):
  """Logs a custom metric in a statsD-esque format to stdout.
 

python/tests/functional/test_logger.py

@@ -80,6 +80,19 @@ def test_setup_service_env_var(monkeypatch, root_logger, stdout):
  assert service_name == log["service"]
 
 
+def test_setup_sampling_rate(monkeypatch, root_logger, stdout):
+ # GIVEN samping rate is explicitly defined via POWERTOOLS_SAMPLE_RATE env
+ # WHEN logger is setup
+ # THEN sampling rate should be equals POWERTOOLS_SAMPLE_RATE value
+ sampling_rate = "0.1"
+ monkeypatch.setenv("POWERTOOLS_SAMPLE_RATE", sampling_rate)
+
+ logger = logger_setup()
+ logger.info("Hello")
+ log = json.loads(stdout.getvalue())
+ assert sampling_rate == log["sampling_rate"]
+
+
 def test_inject_lambda_context(root_logger, stdout, lambda_context):
  # GIVEN a lambda function is decorated with logger
  # WHEN logger is setup
@@ -128,9 +141,8 @@ def handler(event, context):
 
 
 def test_inject_lambda_context_log_event_request_env_var(
- monkeypatch, root_logger, stdout, lambda_context
+  monkeypatch, root_logger, stdout, lambda_context
 ):
-
  # GIVEN a lambda function is decorated with logger instructed to log event
  # via POWERTOOLS_LOGGER_LOG_EVENT env
  # WHEN logger is setup
@@ -160,7 +172,7 @@ def handler(event, context):
 
 
 def test_inject_lambda_context_log_no_request_by_default(
- monkeypatch, root_logger, stdout, lambda_context
+  monkeypatch, root_logger, stdout, lambda_context
 ):
  # GIVEN a lambda function is decorated with logger
  # WHEN logger is setup
@@ -281,12 +293,12 @@ def test_log_metric_multiple_dimensions(capsys):
  "invalid_input,expected",
  [
  (
- {"unit": "seconds"},
- "MONITORING|0|Seconds|test_metric|DemoApp|service=service_undefined\n",
+  {"unit": "seconds"},
+  "MONITORING|0|Seconds|test_metric|DemoApp|service=service_undefined\n",
  ),
  (
- {"unit": "Seconds", "customer": None, "charge_id": "123", "payment_status": ""},
- "MONITORING|0|Seconds|test_metric|DemoApp|service=service_undefined,charge_id=123\n",
+  {"unit": "Seconds", "customer": None, "charge_id": "123", "payment_status": ""},
+  "MONITORING|0|Seconds|test_metric|DemoApp|service=service_undefined,charge_id=123\n",
  ),
  ],
  ids=["metric unit as string lower case", "empty dimension value"],