Advanced Metrics

Advanced metrics in BlazeBee are designed for environments where basic system visibility is insufficient. These collectors expose low-level kernel, hardware, and service-specific signals that are critical for troubleshooting performance degradation, capacity issues, and hardware anomalies.

Advanced collectors are disabled by default and must be explicitly enabled at build time and in configuration.

Purpose

This page describes:

What advanced collectors are available
What subsystems they observe
When and why they should be used
How they integrate into the standard metrics pipeline

These metrics are intended for experienced operators and diagnostic workflows.

Available Advanced Collectors

Thermal

Collector name: thermal

Reads temperature data from thermal zones
Sources include /sys/class/thermal
Reports per-sensor temperatures in degrees

Use cases:

Detect overheating CPUs or SoCs
Monitor passive cooling efficiency
Prevent thermal throttling

Power

Collector name: power

Exposes battery and power-supply information
Reads from /sys/class/power_supply
Includes charge level, voltage, current, and status

Use cases:

Edge devices
Battery-backed systems
Power consumption diagnostics

Pressure (PSI)

Collector name: pressure

Reports Linux PSI (Pressure Stall Information)
Covers CPU, memory, and IO pressure
Indicates how often tasks are stalled due to resource contention

Use cases:

Diagnosing latency spikes
Capacity planning
Identifying hidden resource saturation

Systemd

Collector name: systemd

Queries systemd unit states
Tracks service health and activation status
Requires systemd-based Linux distributions

Use cases:

Service availability monitoring
Detecting failed or flapping units
Infrastructure observability without agents

NTP

Collector name: ntp

Reports clock offset and synchronization state
Uses system time sources
Indicates drift relative to reference clocks

Use cases:

Distributed systems
Time-sensitive workloads
Debugging clock skew issues

Hardware Monitoring (HWMON)

Collector name: hwmon

Reads hardware sensors via /sys/class/hwmon
Includes voltages, fan speeds, temperatures

Use cases:

Bare-metal monitoring
Detecting failing components
Environmental diagnostics

Network Statistics (Extended)

Collector names: arp, netstat, conntrack

ARP cache statistics
Kernel network counters
Connection tracking table usage

Use cases:

Network debugging
Detecting connection leaks
Firewall and NAT diagnostics

File Descriptors

Collector name: filefd

Reports file descriptor limits and usage
Reads from /proc/sys/fs and process tables

Use cases:

Preventing FD exhaustion
Debugging connection-heavy services
Capacity tuning

Additional Advanced Collectors

Depending on build features, BlazeBee may also support:

mdraid — RAID array health
edac — memory error detection
schedstat — scheduler statistics
entropy — kernel entropy pool
filesystem — detailed FS stats
powercap — RAPL power limits

Enabling Advanced Metrics

Build-Time Requirement

Advanced collectors are included only when built with the large feature set or explicit collector features.

cargo build --release --features "blazebee-mqtt-v3 large"

Runtime Configuration

Each advanced collector must be enabled in config.toml:

[[metrics.collectors.enabled]]
name = "pressure"
[metrics.collectors.enabled.metadata]
topic  = "metrics/pressure"
qos    = 1
retain = true

One block per collector
Topic structure is fully user-defined
QoS and retain flags apply per collector

Operational Considerations

Some collectors require elevated privileges
Linux-only functionality
Increased I/O and parsing overhead
Recommended for diagnostic or targeted deployments

Summary

Advanced metrics provide deep visibility into system internals. They are powerful but should be enabled selectively. When used appropriately, they enable early detection of failures, performance bottlenecks, and hardware issues that are invisible to standard monitoring.