promql

To calculate the per-second rate in PromQL, apply the rate() function to a range vector like http_requests_total[5m]. Ensure you include the time window in brackets and aggregate labels after computing the rate to maintain accuracy.

Prometheus histogram_quantile returns incorrect percentiles when the le label is missing from the aggregation. You must group by le inside the sum() function before passing the result to histogram_quantile to compute accurate p95 or p99 latency values.

To optimize slow Prometheus queries caused by high cardinality, use topk() or count() patterns to identify heavy series. Choose scrape-interval-aware window sizes and apply recording rules to pre-compute expensive expressions for dashboard performance.

Create recording rules for SLO calculations by defining PromQL expressions that compute error rates and latency budgets, then saving them as rules in your Prometheus configuration. This pre-computes alert-friendly query patterns for faster dashboard rendering.

Yes, the PromQL queries generated are compatible with Grafana Cloud Metrics and Mimir. The validation covers correct range vectors, aggregation placement, and window sizing to ensure performant time-series expressions across these platforms.

You should always aggregate labels after using rate() in PromQL. Aggregating before applying rate() to a counter produces mathematically incorrect results because it sums counter resets, leading to inaccurate per-second calculations.