Networking Minerva User Guide
Networking Minerva User Guide
Breadcrumbs

Hardware Health

The Hardware Health dashboard provides visibility into physical infrastructure sensor data — component temperatures, PSU output power, voltage, and fan speed — for the following personas.

Persona

Primary Use Case

DC Operations

Monitor power consumption, rail voltages, and thermal readings across all chassis to identify failing PSUs or thermal hotspots before they cause downtime.

Site Reliability Engineers (SRE)

Track trending degradation in voltage rails or fan speed to schedule proactive maintenance rather than reactive replacement.

Hardware Engineering

Validate that chassis run within vendor-specified thermal and voltage envelopes under idle and loaded conditions.

Tenant Admins

Ensure cluster-wide hardware indicators remain nominal over a 24-hour window to support SLA compliance.

Dashboard Overview

The Hardware Health dashboard aggregates Redfish-sourced telemetry from the system_v1 data source to surface physical sensor readings across the cluster. By default, it covers the last 24 hours. Each section targets a distinct hardware subsystem.

Temperature

Monitors thermal state of chassis enclosures and memory components. Both panels display only values above 35°C to surface meaningful thermal events against baseline noise.

  • Chassis Temperature (°C): A heatmap showing the maximum chassis-ambient temperature per device and component over time, filtered to the Chassis subtype. High-density heat concentrated in a single row indicates sustained thermal stress on that enclosure — a signal of airflow restriction, a failed inlet fan, or a failing cooling unit.

  • Memory Temperature (°C): A heatmap showing the maximum temperature across memory components per device and component, filtered to the Memory subtype. Elevated readings on a subset of DIMMs within a node are an early indicator of thermal instability or inadequate cooling in memory bays, especially in GPU-dense nodes with high memory bandwidth.

Output Power

Monitors PSU output power delivered across the cluster, providing visibility into power draw trends and capacity headroom.

  • Average Output Power (kW): A line chart showing the average and maximum PSU output power (in kilowatts) per power supply unit, sampled at 5-minute granularity. Each line represents a distinct PSU identified by its key dimension (device + component). Sustained readings near PSU capacity limits indicate the need for load balancing or hardware review. Sudden drops to near-zero indicate a PSU trip or failure and warrant immediate investigation.

Voltage

Monitors voltage stability across multiple supply rails. Voltage deviations outside the nominal range indicate failing components, aging PSUs, or load-distribution issues. All voltage heatmaps show the minimum reading per device and component and exclude zero values so offline devices do not mask real degradation.

  • Average Voltage (V): A line chart showing average voltage per key (device + component), sampled hourly. This aggregated view surfaces slow-trending rail drift across the cluster that may not be visible in per-rail heatmaps.

  • Min Voltage – 12V Heatmap: A heatmap of minimum 12V rail readings per device and component. The 12V rail powers primary compute and storage components.

  • Min Voltage – 5V Heatmap: A heatmap of minimum 5V rail readings per device and component. The 5V rail supplies lower-power logic components; sustained low readings indicate PSU aging or load imbalance across rails.

  • Min Voltage – 3.3V Heatmap: A heatmap of minimum 3.3V rail readings, matched by component names containing 3.3V. This rail powers embedded controllers and low-power I/O logic; deviations affect baseboard management controller (BMC) stability.

  • Min Voltage – 2.5V Heatmap: A heatmap of minimum 2.5V rail readings, matched to the 2_5V component. This rail is common in memory and I/O subsystems; persistent low readings on a single node are a leading indicator of DIMM or storage controller degradation.

Fan Speed

Monitors cooling fan performance across the cluster. Fan failures or abnormal speed deviations are early indicators of thermal management issues that will manifest as temperature alerts if left unaddressed.

  • Fan Speed (RPM): A line chart showing both the maximum and minimum fan speed (RPM) per key, sampled hourly over the last 24 hours. The spread between max and min for a given device reveals fan imbalance within a chassis. A drop to zero RPM in the minimum line on a device that is otherwise reporting heat indicates a stalled or failed fan.

Last updated: