Metrics Plugin

The Metrics plugin collects system metrics from all available sources with automatic detection. It discovers CPUs, disks, network interfaces, and GPUs without manual configuration.

Binary: logflux-metrics
Platform: Linux, macOS
Entry type: Metric
Service: logflux-agent-metrics.service

Installation

The metrics plugin is included in the LogFlux Agent package and enabled by default. After installing the agent:

1
sudo systemctl enable --now logflux-agent-metrics

Command-Line Flags

Flag	Default	Description
`-config FILE`	–	Path to YAML configuration file
`-interval D`	`60s`	Collection interval (minimum 15s)
`-batch-size N`	`10`	Metrics per batch
`-no-batch`	`false`	Disable batching
`-cpu`	`true`	Collect CPU metrics
`-memory`	`true`	Collect memory metrics
`-disk`	`true`	Collect disk metrics
`-network`	`true`	Collect network metrics
`-system`	`true`	Collect system info
`-gpu`	`true`	Collect GPU metrics (if available)
`-verbose`	`false`	Enable verbose output
`-version`	–	Show version and exit

Configuration File

Default location: /etc/logflux-agent/plugins/metrics.yaml

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
name: metrics
version: 1.0.0
source: metrics-plugin

agent:
    socket_path: /var/run/logflux-agent/agent.sock
    network: unix
    connect_timeout: 10s
    max_retries: 3
    retry_delay: 1s

plugin:
    interval: 60s          # Collection frequency
    collect_cpu: true
    collect_memory: true
    collect_disk: true
    collect_network: true
    collect_system: true

logging:
    level: info
    labels:
        component: metrics
        plugin: metrics
    verbose: false

batch:
    enabled: true
    max_size: 10
    flush_interval: 5s
    auto_flush: true

Collected Metrics

CPU

Per-core user, system, idle, iowait percentages
Load averages (1, 5, 15 minutes)
IRQ and steal time on supported systems

Memory

Total and available physical memory
Swap usage
Cache and buffer sizes
Memory pressure indicators

Disk

Space usage per mounted filesystem
I/O statistics per block device
IOPS and throughput
Virtual filesystems (procfs, sysfs) excluded automatically

Network

Per-interface bytes sent/received
TCP connection states
Error and drop rates
Physical vs virtual interface detection

GPU

Detected automatically when management tools are present:

NVIDIA: nvidia-smi (utilization, memory, temperature)
AMD: rocm-smi
Intel: intel_gpu_top

No configuration needed – if the GPU tools are installed, metrics are collected.

Output Format

Metrics are sent as structured JSON:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
{
    "metric_type": "cpu",
    "timestamp": 1234567890,
    "host": "server01",
    "cpu0_user_percent": 25.5,
    "cpu0_system_percent": 10.2,
    "cpu0_idle_percent": 60.3,
    "cpu0_iowait_percent": 4.0,
    "load_avg_1": 1.24,
    "load_avg_5": 1.15,
    "load_avg_15": 1.08
}

Usage Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
# Default (all metrics, 60s interval)
logflux-metrics

# Faster collection
logflux-metrics -interval 30s

# CPU and memory only
logflux-metrics -disk=false -network=false -gpu=false

# Debugging
logflux-metrics -verbose -interval 5s

Performance

CPU usage: < 0.5% on average
Memory: ~10-20 MB resident
Collection is non-blocking
Batching reduces network overhead

Platform Support

Platform	CPU	Memory	Disk	Network	GPU
Linux	Full	Full	Full	Full	NVIDIA, AMD, Intel
macOS	Full	Full	Full	Full	–

Service Management

1
2
3
4
sudo systemctl start logflux-agent-metrics
sudo systemctl stop logflux-agent-metrics
sudo systemctl status logflux-agent-metrics
sudo journalctl -u logflux-agent-metrics -f