Only 1 watchtower handler works; others fail when logging with "WatchtowerWarning: Received message after logging system shutdown"

[ebk46]

We are using Django 4.2.15 and Watchtower 3.3.1

We have logging via Watchtower for 2 different CloudWatch streams within the same group. The following is environment-specific settings code within -settings.py files. They share a common initial LOGGING config.

boto3_logs_client = boto3.client("logs", region_name=AWS_REGION)

LOGGING["handlers"]["primary"] = {
  "class": "watchtower.CloudWatchLogHandler",
  "level": "DEBUG",
  "boto3_client": boto3_logs_client,
  "log_stream_name": "primary",
  "formatter": "aws",
}

LOGGING["loggers"]["primary"] = {
  "handlers": ["primary"],
  "level": "DEBUG",
}

LOGGING["handlers"]["secondary"] = {
  "class": "watchtower.CloudWatchLogHandler",
  "level": "DEBUG",
  "boto3_client": boto3_logs_client,
  "log_stream_name": "secondary",
  "formatter": "aws",
}

LOGGING["loggers"]["secondary"] = {
  "handlers": ["secondary"],
  "level": "DEBUG",
}

The "primary" logger works as intended. However, any logs to the "secondary" logger fail with:

WatchtowerWarning: Received message after logging system shutdown

As with many issues in this repo, setting use_queues: False enables logs for both, but that is not suitable for a production deployment.

[ebk46]

It seems that switching our gunicorn worker class from the default ("sync") to "gthread" and using multiple threads per worker somehow resolved this. I'm not sure exactly how, but I'll close this issue since it seems to not be directly related to watchtower. With that said, if that's true, it would probably be a helpful addition to the docs.