Signal Suppression

What is a signal?

Before a detection becomes an incident, it passes through a suppression window. During this window, DataAgent waits to see if the condition persists across multiple analysis cycles.

This prevents transient events — a pod briefly restarting during a rolling update, a node momentarily under pressure during a scale event — from generating incidents.

How suppression works

1. K8sGPT detects a condition → a signal is created with status pending
2. The signal waits through the suppression window
3. If the condition is still present when the window expires → signal is promoted to an incident
4. If the condition disappears within the window → signal is suppressed (no incident created)

Suppression window

The default suppression window is configured at the organisation level. Contact support to adjust it.

A shorter window reduces latency to incident creation but increases noise from transient conditions. A longer window reduces noise but delays incident detection.

Manually suppressed signals

Signals can be suppressed by name-based suppression rules. If a resource regularly triggers false-positive detections, contact support to add a suppression rule for that resource.

Viewing signal state

Signal state is not exposed directly in the dashboard. If you believe a genuine incident is being suppressed, check Incidents not appearing.