Prometheus Query Builder

PromQL (Prometheus Query Language) is the functional expression language used to query Prometheus metrics. It lets you select and aggregate time series data in real time. PromQL queries return instant vectors, range vectors, scalars, or strings.

Label matchers filter which time series are selected. The four operators are: = (exact match), != (not equal), =~ (regex match), and !~ (regex not match). For example, {job=\"api-server\", status=~\"5..\"} selects series where job is api-server and status matches 5xx.

rate() calculates the per-second average rate over the full range window, smoothing out spikes. irate() uses only the last two data points, making it more responsive to sudden spikes but noisier. Use rate() for most alerting and irate() for burst detection.

Use increase() when you want the total count of events over a time window. increase(http_requests_total[1h]) gives requests in the last hour; rate(http_requests_total[5m]) gives requests/sec averaged over 5 minutes.

by() keeps only listed labels and aggregates everything else together. without() drops listed labels and keeps all others. They are inverses of each other. Use by(job) for per-job totals; use without(instance) to remove high-cardinality labels.

Use histogram_quantile(phi, rate(metric_bucket[range])). For example: histogram_quantile(0.95, rate(http_request_duration_seconds_bucket[5m])) gives the p95 latency. The metric must be a Prometheus histogram type with _bucket suffix.