Observability — cheat sheet
A one-page lookup for the terms used across this course. Grouped by task; the course pages cover the
why behind each one — this page is just the what.
Three pillars & golden signals
| Pillar |
Answers |
Nature |
| Metrics |
Is something wrong, in aggregate, right now? |
A number over time |
| Logs |
What exactly happened, at one point in time? |
A discrete event |
| Traces |
Where did the time go, across services, for one request? |
A request's path, timed hop by hop |
| Golden signal |
Question |
| Latency |
How long do requests take? |
| Traffic |
How much demand is the service under? |
| Errors |
How much of that traffic is failing? |
| Saturation |
How full is a constrained resource? |
Prometheus & PromQL
| Term |
What it means |
| Scrape |
Prometheus pulls a target's /metrics endpoint on a fixed interval |
| Exporter |
A standalone process that re-exposes a third-party thing's stats at /metrics |
| Counter |
A metric that only ever goes up (or resets to 0 on restart) |
| Gauge |
A metric that goes up or down freely |
| Histogram / Summary |
A metric bucketed to derive quantiles/averages |
| Instant vector |
A metric's latest value (e.g. http_requests_total) |
| Range vector |
A metric's values over a window (e.g. http_requests_total[5m]) |
rate(metric[5m]) |
Per-second average over the last 5 minutes |
Grafana
| Term |
What it is |
| Data source |
The backend Grafana queries (Prometheus, a logs/traces store, ...) |
| Panel |
One graph/number/table/heatmap built from a data-source query |
| Dashboard |
A collection of panels arranged on one page |
| Task |
Mechanism |
| Confirm what an alert already told you |
Open the dashboard for more context, not the other way round |
Alerting, SLOs & error budgets
| Term |
What it means |
| Alerting rule |
A PromQL expression evaluated on a schedule; fires when true for a configured duration |
| Symptom-based alert |
Tests something user-facing (a golden signal) — the default choice |
| Cause-based alert |
Tests an internal condition that might lead to user pain |
| Alertmanager |
Dedupes, groups, routes, and silences alerts a rule has fired |
| SLI |
The actual measured number for a service |
| SLO |
A target for that SLI over a time window |
| Error budget |
The gap between 100% and the SLO — the failure allowed before violating it |
How this connects
The three pillars & golden signals page covers what to watch and why, the Prometheus & PromQL page covers
how metrics get collected and queried, the Grafana dashboards page covers turning a query into something a
human looks at, and the alerting, SLOs & error budgets page covers turning a crossed threshold into a
notification and a reliability target into a number.