project-charlie/.github/skills/charlie-grafana-logs/SKILL.md

name, description

name	description
charlie-grafana-logs	Fast path for checking Grafana/VictoriaLogs logs for Charlie stacks and containers. Use when asked to inspect logs, errors, container output, or symptoms for a Charlie service.

Charlie Grafana logs

Use the Loki-compatible Grafana datasource first:

• Datasource UID: victorialogs-loki
• Datasource name: VictoriaLogs Loki
• Common labels: compose_project, service_name, service, container, host, job, stream

Preferred flow:

1. If the stack is known, query by compose_project="<stack>".
2. If the service/container is known, add service_name="<service>".
3. Use query_loki_stats on the selector before pulling logs.
4. Pull a small recent sample with query_loki_logs.
5. For errors, use case-insensitive filters for terms like error, fail, fatal, panic, denied, permission, warning, unable, and cannot.
6. Use count_over_time for exact counts when the frequency matters.

For Docker Compose stacks, the most precise selector is usually {compose_project="<stack>", service_name="<service>"}.

If labels are missing or the selector returns no streams, list label names and values for the same time range before changing the query shape.

Capture the datasource, selector, time range, count, and representative log lines in troubleshooting notes so the same evidence can become a TSG or skill update.