NVIDIA Minerva includes a library of preconfigured alerts that identify hardware degradations, network connectivity issues, and application performance bottlenecks. These alerts are optimized for the high-scale requirements of NVIDIA AI factories, ensuring you are notified of critical events before they impact training jobs or cluster availability.
Multi-Channel Notification Support
Preconfigured alerts are compatible with the following notification channels:
-
Slack: richly formatted alerts with direct links to relevant dashboards.
-
Webhooks: customizable payloads for integration with third-party systems.
Preconfigured Alert Templates
|
Alert Template |
Description |
|---|---|
|
GPU Temp High |
Triggers when GPU core temperature exceeds thresholds (see GPU Summary by GPU). |
|
NVLink BW Loss |
Triggers when NVLink bandwidth deviates from the established baseline for a specific node or switch. |
|
Rare Log with Remediation |
Uses AI-driven log analysis to surface uncommon logs and provide AI-generated remediation steps. |
|
Collective Learning |
Uses supervised machine learning to parse syslogs, identify critical patterns, and automatically classify them into specific |
Last updated: