Error logging - moving away from email

[ArneTR]

Since we already have a file logging implemented in the https://github.com/green-coding-berlin/green-metrics-tool/blob/main/lib/error_helpers.py we should:

• Define a file where to log runner errors
• Define a file where to log api errors (My suggestion would be to bind a writeable file into the container as bind-mount)
• The user then can define a transport mechanism of choice, for instance pushing all logs into a logstash etc.

Currently every failure in the API takes a lot of processing time. So even pushing to a remote service would create quite some load.

The filesystem solution decouples it completely.

Our implementation to have then again email, but async, could be using swatch

[ribalba]

What about we implement https://docs.python.org/3/library/logging.html . This is something I have had on my todo for some time. We could implement logging for the runner and the server and such have a unified api. We could also start the server/ runner in debug mode to see what is happening, if we need to and otherwise reduce the "clutter" a little.

[ArneTR]

yes, can be done. We can just swap it for the standard library to make it more semantic / obvious for frequent Python users.

The logging is already centralized and unified in server and CLI though, so from a functional perspective I see no uptake ... but still agree to do it once we tackle it.