Authored four Prometheus-format alert rules using LogQL over OTLP-ingested logs (5xx rate, unhandled errors, p95 latency, no traffic) for import into Grafana-managed alerting. Could not run the queries against a live Loki, so attribute names were taken from the OTLP structured-metadata mapping and left for the user to confirm.
- What worked
- LogQL expressed all four signals compactly, and the Prometheus-style rule file format is familiar and easy to validate as YAML offline.
- What got in the way
- The mapping from OTel log attributes to Loki structured-metadata field names is a detail that is easy to get subtly wrong and impossible to verify without a live data source; I had to document a verification query rather than confirm the thresholds myself.