Error logging is incompatible with Lambda Advanced Logging Controls #202

Activity

1. anzheyazzz commented on Jul 17, 2026

   

   anzheyazzz

   on Jul 17, 2026

   Contributor

   More actions

   Hi! Thanks for raising the issue.
   I did a quick test with managed runtime: python3.14

   import logging
   
   def lambda_handler(event, context):
       logger = logging.getLogger()
       logger.error('Error testing message')
       logger.info('Info testing message')
   
       a = []
       print(a[1])
   
       return {'statusCode': 200}

   and I can see from CloudWatch, the error was correctly filtered:

   

   My setting:

   

   Can you share a bit more about how you setup the filtering?
2. maoyouaa commented on Aug 10, 2026

   

   maoyouaa

   on Aug 10, 2026

   More actions

   I investigated this further and prepared a small patch.

   The managed Python 3.14 result may be explained by the log sink used. FramedTelemetryLogSink emits errors with an ERROR frame, while the stdout fallback, StandardLogSink, writes JSON containing "log_level": "ERROR" but no "level" field.

   The patch adds "level": "ERROR" while retaining "log_level" for compatibility, with a regression test. The test fails on current main (3810eed) and passes with the patch; make pr passes all 133 tests with 95.72% coverage.

   I have not reproduced this end-to-end in CloudWatch or confirmed that deployed Lambda environments can use the stdout fallback. If aligning that schema is still useful, I’m happy to submit a PR.
3. maxday commented on Sep 14, 2026

   

   maxday

   on Sep 14, 2026

   Member

   More actions

   closing as #220 has been merged
4. maxday

   closed this as completedon Sep 14, 2026