Severity Levels
How DataAgent classifies incident severity.
Every incident is assigned one of three severity levels by K8sGPT at detection time.
Levels
Critical
Service degradation or failure is occurring now. Immediate action is required.
Warning
Elevated risk. The system is degraded or trending toward failure. Action is needed soon.
Info
A condition worth monitoring. No immediate action required.
Examples by severity
Critical
- Pod in
CrashLoopBackOffwith no successful starts - Pod killed by OOM (out-of-memory) killer
- Node in
NotReadystate - Node under
MemoryPressureorDiskPressure - StatefulSet pod unable to bind its PersistentVolumeClaim
- Deployment with zero available replicas
Warning
- Pod with high restart count (not yet in crash loop)
- Pod stuck in
Pendingstate (scheduling constraint) - Image pull failure
- Deployment with reduced availability (some but not all replicas ready)
- Node under
PIDPressure - CronJob that missed its scheduled run
- Job that failed within its backoff limit
Info
- Workload with availability below configured threshold but above critical floor
- Resource usage trending toward limit
- K8s warning events without current impact
Filtering by severity
Use the severity filter in the Incidents view to focus on the level that matters most. You can combine severity with status and cluster filters.