ReadMeFeed / Community knowledge
Real questions. Useful conversations. Find the people who know your stack.
DevOps & observability · #prometheus
Learn how to offload expensive PromQL expressions into recording rules, the trade‑offs involved, and how to verify the rule works as expected.
Learn how Grafana Explore turns ad-hoc metric debugging into a fast, repeatable workflow with a concrete Prometheus example.
Learn how adjusting the evaluation_interval of Grafana alert rules balances detection latency against query load, with a concrete example and practical verification steps.
Stop copying dashboards per environment. Grafana templating variables let one dashboard serve every region and service — here's how chained query variables work, with the performance catch to watch for.
Drop high‑cardinality labels at scrape time with Prometheus metric_relabel_configs. Learn how to prune labels, verify the impact, and understand the trade‑offs before you hit the TSDB limits.
Recording rules precompute expensive PromQL into new time series so dashboards and alerts stop re-scanning raw samples. A worked example, verification steps, and the cardinality trade-off.
Prometheus rejects out-of-order samples when a series timestamp arrives earlier than the newest stored point. Find the offending series, check NTP, remove explicit timestamps from client code, and use the out-of-order window as a grace period, not a fix.
Dashboards re-running heavy PromQL on every refresh? Recording rules precompute expensive expressions into new series. A worked example, verification steps, and the real costs.
Rancher's built-in monitoring stack gives you Prometheus and Grafana per cluster with a single toggle. This post walks through enabling it on an RKE2 cluster, verifying the deployment, and understanding where the defaults fall short.
grafana
rancher