PromQL Cheatsheet
Top patterns.
Overview
PromQL is Prometheus’s query language for time-series metrics. Fluency at PromQL is the difference between dashboards that answer questions and dashboards that look pretty; the patterns are well-known and worth memorising.
- Label selectors.
http_requests_total{status="500"}; narrow scope by labels; the foundation of every query. - Range vectors.
[5m]selects a 5-minute window; the input to rate calculations and aggregations over time. - rate() and irate(). Per-second rate of change; turn counters into the rate metrics that dashboards actually want.
- Aggregation plus histogram_quantile().
sum,avg,maxfor summaries;histogram_quantilefor SLO percentiles.
The approach
The practical approach: rate of counters as the default, histogram_quantile for percentiles, sum by for aggregation. The team’s discipline produces fast queries that survive cardinality growth.
- rate(http_requests_total[5m]). Request rate per second over 5-minute window; the canonical traffic query.
- sum by(service) (rate(...)). Per-service total rate; the aggregation pattern that powers most dashboards.
- histogram_quantile(0.99, sum by(le) (rate(...))). p99 latency from histogram buckets; the SLO query template.
- topk(10, ...). Top 10 series; fast attribution during investigation; "which service is the offender?"
- increase(counter[1h]). Total counter increase over an hour; the volume query template; supports capacity planning.
Why this compounds
PromQL fluency compounds across queries. Each query teaches the data model; the team’s metric query speed accrues; cost-per-question falls as the muscle matures.
- Faster dashboards. Fluent PromQL produces fast dashboards; the panel renders in milliseconds, not seconds.
- Better alerting. Mature queries produce reliable alerts; false positives drop; on-call trust is preserved.
- Cost discipline. Cardinality-aware queries reduce Prometheus load; the bill stays linear, not exponential.
- Institutional knowledge. Each query teaches Prometheus’s data model; the team’s observability muscle grows.
PromQL fluency is an operational discipline that pays off across years. Nova AI Ops integrates with Prometheus, surfaces patterns, and supports the team’s monitoring discipline.