Debug logs duplication due to incorrect configuration

[nkhitrov]

How do you use Sentry?

Self-hosted/on-premise

Version

2.19.2

Steps to Reproduce

Sentry duplicates log records when you set debug mode via sentry_sdk.init() and configure handler in root logger only.

In my project I use structlog so full code example with it

import logging
import sys

import structlog

LOG_LEVEL = logging.DEBUG


def remove_processors_meta(
    _,
    __,
    event_dict,
):
    del event_dict["_record"]
    del event_dict["_from_structlog"]
    event_dict.pop("positional_args", None)
    return event_dict


structlog.configure_once(
    processors=[
        structlog.stdlib.ProcessorFormatter.wrap_for_formatter,
    ],
    logger_factory=structlog.stdlib.LoggerFactory(),
    wrapper_class=structlog.stdlib.BoundLogger,
    cache_logger_on_first_use=True,
)
formatter = structlog.stdlib.ProcessorFormatter(
    pass_foreign_args=True,
    processors=[
        structlog.stdlib.add_logger_name,
        structlog.stdlib.add_log_level,
        structlog.processors.TimeStamper(fmt="%Y-%m-%d %H:%M:%S", utc=True),
        structlog.dev.set_exc_info,
        remove_processors_meta,
        structlog.dev.ConsoleRenderer(
            event_key="message",
            exception_formatter=structlog.dev.better_traceback,
        ),
    ],
)

handler = logging.StreamHandler(stream=sys.stdout)
handler.setFormatter(formatter)

root_logger = logging.getLogger()
root_logger.addHandler(handler)
root_logger.setLevel(LOG_LEVEL)


l = logging.getLogger(__name__)
l.info("test1", extra={"extra": "one"})

import sentry_sdk  # must be after configuration of logging and structlog
from sentry_sdk.utils import logger

sentry_sdk.init("https://123@localhost/1", debug=True)

Expected Result

Sentry does not write log records via own stream handler and use parent (root) logger configuration

Output:

2025-01-18 06:44:33 [info     ] [__main__] event=test1
2025-01-18 06:44:33 [debug    ] [sentry_sdk.errors] event=Setting up integrations (with default = True)
2025-01-18 06:44:33 [debug    ] [sentry_sdk.errors] event=Did not import default integration sentry_sdk.integrations.anthropic.AnthropicIntegration: Anthropic not installed

Actual Result

Output:

 [sentry] DEBUG: Setting up integrations (with default = True)
2025-01-18 06:11:54 [debug    ] Setting up integrations (with default = True) [sentry_sdk.errors] func_name=setup_integrations module=__init__ pathname=/<...>/.venv/lib/python3.11/site-packages/sentry_sdk/integrations/__init__.py process=69509 process_name=MainProcess thread=8643448640 thread_name=MainThread

Possible fix

In debug.py we should change configuration check and use special logger method to check parent handlers too

from this

def init_debug_support():
    # type: () -> None
    if not logger.handlers:
        configure_logger()

to this

def init_debug_support():
    # type: () -> None

    if not logger.hasHandlers():  # check parent handlers
        configure_logger()

[szokeasaurusrex]

Good idea @nkhitrov

[sentrivana]

Thanks @nkhitrov for the proposed solution, too -- added it. The fix will be out with the next SDK release, probably next week.

[szokeasaurusrex]

Hi @nkhitrov – just writing to let you know that we are going to revert your proposed fix (#4133) because it caused a bug (#4266) where folks with custom log configurations were unexpectedly getting a bunch of debug logs from the Sentry SDK. The change in behavior was due to #4133, and we realized at that point that we in fact did not want the SDK to consider parent loggers when setting up the SDK logger.

We are going to need to come up with a more robust solution for your use case. I will reopen your issue.