Live Metrics is the streaming view for the metrics that matter most during an incident. Sub-second updates, side-by-side comparison, single-service deep-dive. Open it during a war room when polling-based dashboards feel sluggish; close it when the incident is over and you go back to the normal dashboards.
Live Metrics uses a WebSocket stream from the metrics backend, not periodic HTTP polling. Updates land within a second of the metric being written. The connection multiplexes multiple metrics over one socket so opening 10 charts does not create 10 connections. Reconnect is automatic on network blips.
During an incident, comparing the affected service to a healthy peer is often the fastest diagnosis. Side-by-side mode renders two services' live metrics in the same chart with synchronized scales. The visual difference is the diagnosis: "payments p95 climbing while orders is flat means it is not a shared upstream."
When you open an incident page, the relevant live metrics auto-pin to the war-room view. The pinned charts cover the affected service's top SLIs. The data is captured into the incident archive at close time so the postmortem reader sees the live charts as they were during the incident.
Streaming is a more expensive query pattern than periodic polling. Each open chart costs roughly 5x what the equivalent dashboard widget costs. The page is built for incidents, not always-on viewing. Inactive sessions auto-close after 30 minutes of no interaction. Close the page when the incident is over.
Subscribe to Nova AI Ops on YouTube for demos, tutorials, and feature deep-dives.
Live Metrics is built for the war room. Sub-second updates, no polling lag, no waiting for a 30-second refresh.