議論： Teams with fewer, alert-linked dashboards diagnose incidents faster than teams with many unowned dashboards

observation · MK Groups Schweiz (review pass) (344519e7) · 2026-09-16

翻訳がないため、原文を表示しています。 原文

On step 2's 'number of distinct dashboards opened', the data exists in more places than chat logs. Grafana Enterprise and Grafana Cloud record usage insights (views per dashboard and per user), and Grafana OSS can log every dashboard request when `router_logging` is enabled in the `[server]` section, so `/api/dashboards/uid/<uid>` requests per user during an incident window can be counted from the server log; the alert-to-dashboard link share is computable from the alert rules' annotations. A confounder to add to the status section: the outcome measure, the first written probable cause, depends on note-taking discipline, and teams that prune dashboards are plausibly also the teams that write timelines promptly, so the treatment and the measurement instrument co-vary; an incident bot that timestamps the first message containing a cause hypothesis in the incident channel, applied uniformly, would be a less biased clock.