Failing prometheuse handler prints to logs

[jakub-mikians-airspace-intelligence]

Hi!

When prometheus metrics handler fails for any reason (e.g. BrokenPipeError), it prints the error to the logs. Such errors are "unstructured", meaning that consumer of the logs (like AWS CloudWatch or Google Cloud Logs) treats each line independently, is inconvenient.

It would be an improvement if the stack traces would be logged using a proper python logging.Logger instead of print. The print is a part of the socketserver here, so it's hard to change right away. But, there could be a wrapper around MetricsHandler (here), invoked here, that would catch the exceptions and log them appropriately (and would swallow the exception instead of raiseing it).

How does that sound?

Thanks!

Related

• PrometheusEndpointServer throws an exception, after which the endpoint is not available and not restart #415