Operational Best Practices
Dashboard Usage
- Confirm time range, variables, and latest sample timestamp.
- Compare a known-good peer before interpreting an absolute value.
- Use raw queries to distinguish missing data from a dashboard-filter problem.
- Preserve dashboard links with filters and time range in incident notes.
Alert Response
- Distinguish service alerts from scrape/telemetry alerts.
- Inspect alert-instance labels; one rule may represent many targets.
- Preserve the original state before acknowledging, silencing, or editing a rule.
- Verify notification delivery separately from alert evaluation.
Log Analysis
- Start with the expected regional instance, index, and narrow time range.
- Add filters one dimension at a time.
- Use
unknown_syslogto identify enrichment gaps. - Do not treat no search results as proof of no service event.
Incident Investigation
Correlate at least two evidence paths when customer impact is possible: metrics plus logs/traps, or service evidence plus target/scrape health. State clearly when a known coverage gap limits the conclusion.
Maintenance Windows
Identify affected targets, dashboards, rules, and ingress paths before maintenance. Use narrowly scoped, time-bounded silences and verify telemetry recovery before ending the window.