Bug: POWERTOOLS_METRICS_DISABLED is ignored by single_metric, the cold start metric, add_metric auto-flush and Datadog's lambda_metric

Ouverte
  • Quelque chose ne va pas ?
#8,667 2 commentaires 0 réactions 1 personne assignée Voir sur GitHub

Les mainteneurs répondent en général sous 1 jour

@soujasK y travaille déjà.

Depuis le 25/10/2026.

  • #8668 par @soujasK — ouverte

Évaluation

Cette issue n'a pas encore été évaluée.

Description

bug metrics
Expected Behaviour

With POWERTOOLS_METRICS_DISABLED=true (or POWERTOOLS_DEV=true without POWERTOOLS_METRICS_DISABLED), nothing is emitted, as documented ("Disables all metrics emitted by Powertools", for both the CloudWatch and Datadog providers, per #6046).

Current Behaviour

flush_metrics() checks is_metrics_disabled(), but three other emission paths don't:

  1. single_metric (metrics/base.py) always prints in its finally block. This is also how the CloudWatch provider emits the ColdStart metric, so @metrics.log_metrics(capture_cold_start_metric=True) still prints a ColdStart EMF line while the application metrics are correctly suppressed.
  2. The CloudWatch add_metric auto-flush: when the metric set reaches 100 metrics, or a metric reaches 100 values, AmazonCloudWatchEMFProvider.add_metric prints the EMF blob directly.
  3. Datadog with the datadog_lambda package installed: DatadogProvider.flush_metrics calls lambda_metric(...) without checking is_metrics_disabled(); only the flush-to-log branch checks it.

In Lambda the EMF lines in 1 and 2 are ingested as real CloudWatch metrics, so metrics are published even though they're disabled.

While there: if the body of a with single_metric(...) block raises, metric_set is still None in the finally block and a literal null line is printed to stdout.

Code snippet
import os
from types import SimpleNamespace

os.environ["POWERTOOLS_METRICS_DISABLED"] = "true"

from aws_lambda_powertools import Metrics
from aws_lambda_powertools.metrics import single_metric

metrics = Metrics(namespace="ns", service="svc")


@metrics.log_metrics(capture_cold_start_metric=True)
def handler(event, context):
    metrics.add_metric(name="Booking", unit="Count", value=1)  # correctly not emitted


context = SimpleNamespace(function_name="fn", memory_limit_in_mb=128, invoked_function_arn="arn", aws_request_id="id")
handler({}, context)  # prints the ColdStart EMF line

with single_metric(name="Single", unit="Count", value=1, namespace="ns"):
    pass  # prints the EMF line

for i in range(100):
    metrics.add_metric(name=f"metric_{i}", unit="Count", value=1)  # auto-flush prints the EMF blob

Output (develop @ 5785377, Python 3.11):

{"_aws":{"Timestamp":...,"CloudWatchMetrics":[{"Namespace":"ns","Dimensions":[["function_name","service"]],"Metrics":[{"Name":"ColdStart","Unit":"Count"}]}]},"function_name":"fn","service":"svc","ColdStart":[1.0]}
{"_aws":{"Timestamp":...,"CloudWatchMetrics":[{"Namespace":"ns","Dimensions":[[]],"Metrics":[{"Name":"Single","Unit":"Count"}]}]},"Single":[1.0]}
{"_aws": {"Timestamp": ..., "CloudWatchMetrics": [{"Namespace": "ns", "Dimensions": [["service"]], "Metrics": [{"Name": "metric_0", "Unit": "Count"}, ...
Possible Solution

Check is_metrics_disabled() at each of those points, like flush_metrics() does:

  • single_metric: only print when metric_set is not None and not is_metrics_disabled() (which also stops the null line).
  • CloudWatch add_metric: skip the print on auto-flush when disabled, but still clear metric_set as today.
  • Datadog: skip the lambda_metric submission when disabled.

Happy to send a PR with tests for each path.

Steps to Reproduce

Run the snippet above.

Powertools for AWS Lambda (Python) version

latest (develop @ 5785377)

AWS Lambda function runtime

3.11

Packaging format used

PyPi

Debugging logs

No response

Langage dominant
Python
Étoiles
3.3k
Forks
511
Merge moyen
3 j 4 h
PR mergées (30 j)
98

Préparer son environnement

Par où commencer

  1. Lisez l'issue en entier, puis le guide de contribution du projet.
  2. Signalez en commentaire que vous la prenez — cela évite que deux personnes fassent le même travail.
  3. Forkez le dépôt et travaillez sur une branche.
  4. Ouvrez une pull request qui référence le numéro de l'issue.

Autres issues de aws-powertools/powertools-lambda-python

Toutes les issues de aws-powertools/powertools-lambda-python

Issues similaires

Plus d'issues Python

Recevez les nouvelles issues par e-mail

Un résumé court des issues GitHub adaptées aux débutants.