42-transcendence

Node Exporter Documentation

Introduction

Node Exporter is a Prometheus exporter for hardware and OS metrics exposed by *NIX kernels. It provides detailed insights into system-level resources including CPU, memory, disk, network, and more. In our architecture, Node Exporter serves as the primary source of infrastructure metrics, enabling effective resource monitoring and capacity planning.

Configuration in Docker Compose

We’ve configured Node Exporter as follows:

node-exporter:
  <<: *common
  image: prom/node-exporter:latest
  container_name: node-exporter
  profiles: ["grafanaprofile"]
  ports:
    - "9101:9100"
  logging:
    driver: gelf
    options:
      gelf-address: "udp://${LOG_HOST}:12201"
      tag: "node-exporter"
  healthcheck:
    test: ["CMD-SHELL", "wget -q --spider http://localhost:9100/metrics || exit 1"]
    interval: 30s
    timeout: 10s
    retries: 5

Key Configuration Aspects

• Official Image: Uses the official prom/node-exporter image for reliable metrics collection
• Port Mapping: Exposes port 9100 (mapped to 9101 on the host) for metrics access
• Profiles: Part of the “grafanaprofile” group for selective startup
• Logging: Integrates with our ELK stack via GELF
• Healthcheck: Uses wget instead of curl since the Node Exporter container doesn’t have curl installed, checking that the metrics endpoint is functioning properly

Prometheus Integration

Node Exporter is configured as a scrape target in Prometheus:

scrape_configs:
  - job_name: "node"
    static_configs:
      - targets: ["node-exporter:9100"]

This allows Prometheus to automatically collect all system metrics exposed by Node Exporter.

Available Metrics

Node Exporter exposes hundreds of metrics across several categories:

Category	Description	Example Metrics
CPU	Processor usage and statistics	node_cpu_seconds_total, node_load1/5/15
Memory	RAM usage and allocation	node_memory_MemAvailable_bytes, node_memory_MemFree_bytes
Disk	Storage usage and I/O	node_filesystem_free_bytes, node_disk_io_time_seconds_total
Network	Interface traffic and errors	node_network_receive_bytes_total, node_network_transmit_errors_total
System	Uptime, process counts	node_time_seconds, node_procs_running

Grafana Dashboard Integration

For optimal visualization, we use the official Node Exporter dashboard in Grafana (Dashboard ID: 1860). This dashboard provides comprehensive views of:

1. System Load: CPU usage, load averages, and context switches
2. Memory Usage: Available memory, cached memory, and swap usage
3. Disk Space: Filesystem usage across all mounted volumes
4. Disk I/O: Read/write operations and latency
5. Network Traffic: Bandwidth usage and error rates

To import this dashboard:

1. Navigate to Grafana → Dashboards → Import
2. Enter the dashboard ID: 1860
3. Select your Prometheus data source
4. Click Import

Accessing Node Exporter Metrics Directly

You can access raw metrics at:

http://localhost:9101/metrics

This endpoint returns all available metrics in Prometheus format, useful for debugging or exploring available metrics.

Customizing Node Exporter

While we use the default configuration, Node Exporter can be customized with various flags:

command:
  - '--collector.filesystem.ignored-mount-points=^/(sys|proc|dev|host|etc)($|/)'
  - '--collector.textfile.directory=/node_exporter/textfile_collector'

Common customizations include:

• Ignoring specific filesystems
• Enabling/disabling specific collectors
• Adding textfile collector for custom metrics

Security Considerations

In production environments, consider:

• Restricting network access to the Node Exporter port
• Using TLS for secure metrics collection
• Running with reduced privileges

Troubleshooting

If metrics aren’t appearing in Prometheus:

1. Check Node Exporter is running:

   docker compose ps node-exporter
   

2. Verify metrics endpoint is accessible:

   curl http://localhost:9101/metrics
   

3. Check Prometheus targets status:

   http://localhost:9090/targets
   

4. Review Node Exporter logs:

   docker compose logs node-exporter