The monitoring guide has been rewritten to match how logs are collected today, so the setup it describes now works as written.
- Structured JSON logging takes a single configuration property, with no extra library and no XML file to maintain. - The guide lists the labels your log lines actually carry, and warns against filtering on the automatically guessed severity: on a live namespace it labelled none of the real error lines as errors, so an alert built on it can stay silent through an incident. - The sample dashboard has been replaced. The previous one filtered on fields that no longer exist, so it stayed empty for anyone who followed the guide.