Monitor your VPS with node_exporter and Grafana dashboards

To monitor a VPS with node_exporter and Grafana you install the exporter on the server, let Prometheus scrape its metrics endpoint every few seconds, and point Grafana at Prometheus for graphs and alerts. This guide does all three on one Linux VPS, keeps every service bound to localhost, and leaves you with a dashboard that shows CPU, memory, disk and network history.

You will install node_exporter as a systemd service, run Prometheus and Grafana in Docker Compose, and reach both through an SSH tunnel so nothing new is exposed to the internet.

Before you start

• A Linux VPS running Ubuntu 22.04/24.04, Debian 12, or AlmaLinux/Rocky 8–9, with root or sudo access.
• Docker and the Compose plugin installed. If you have not done this yet, follow Deploy a Self-Hosted App with Docker Compose from Scratch first.
• About 1 GB of free RAM and a few GB of disk for metric history. Prometheus stores samples on disk, so plan for growth.
• Your server IP and SSH access. Log in to your VPS provider's control panel if you need to check the IP, reset the root password or open the browser console.

Install node_exporter on the VPS

Download the current release rather than pinning a version you read in a blog post. The releases page on GitHub always lists the latest tag.

cd /tmp
VER=$(curl -fsSL https://api.github.com/repos/prometheus/node_exporter/releases/latest | grep -oP '"tag_name": "v\K[^"]+')
curl -fsSLO "https://github.com/prometheus/node_exporter/releases/download/v${VER}/node_exporter-${VER}.linux-amd64.tar.gz"
tar xzf "node_exporter-${VER}.linux-amd64.tar.gz"
sudo install -m 0755 "node_exporter-${VER}.linux-amd64/node_exporter" /usr/local/bin/node_exporter

On an ARM VPS, replace linux-amd64 with linux-arm64 in both the download and the extracted directory name. Confirm the binary runs:

node_exporter --version

Create a service account with no shell and no home directory, then add the systemd unit.

sudo useradd --no-create-home --shell /usr/sbin/nologin node_exporter
sudo tee /etc/systemd/system/node_exporter.service >/dev/null <<'EOF'
[Unit]
Description=Prometheus Node Exporter
After=network-online.target

[Service]
User=node_exporter
Group=node_exporter
Type=simple
ExecStart=/usr/local/bin/node_exporter --web.listen-address=127.0.0.1:9100
Restart=on-failure
RestartSec=5

[Install]
WantedBy=multi-user.target
EOF
sudo systemctl daemon-reload
sudo systemctl enable --now node_exporter

Check the endpoint returns metrics:

curl -s http://127.0.0.1:9100/metrics | head -n 20

You should see lines such as node_cpu_seconds_total and node_memory_MemAvailable_bytes. If the service fails, sudo journalctl -u node_exporter -n 50 shows why.

Run Prometheus and Grafana with Docker Compose

Create a working directory and the scrape configuration. Prometheus needs to know where node_exporter lives; since both run on the same host, 127.0.0.1:9100 is correct.

sudo mkdir -p /opt/monitoring
cd /opt/monitoring
sudo tee prometheus.yml >/dev/null <<'EOF'
global:
  scrape_interval: 15s
  evaluation_interval: 15s

scrape_configs:
  - job_name: node
    static_configs:
      - targets: ['127.0.0.1:9100']
EOF

Now the Compose file. Both containers use host networking so they can reach the exporter on the loopback address, and both listen only on 127.0.0.1.

services:
  prometheus:
    image: prom/prometheus:latest
    restart: unless-stopped
    network_mode: host
    command:
      - --config.file=/etc/prometheus/prometheus.yml
      - --storage.tsdb.path=/prometheus
      - --storage.tsdb.retention.time=30d
      - --web.listen-address=127.0.0.1:9090
    volumes:
      - ./prometheus.yml:/etc/prometheus/prometheus.yml:ro
      - prometheus-data:/prometheus

  grafana:
    image: grafana/grafana:latest
    restart: unless-stopped
    network_mode: host
    environment:
      GF_SECURITY_ADMIN_PASSWORD: ${GF_ADMIN_PASSWORD}
      GF_SERVER_HTTP_ADDR: 127.0.0.1
      GF_SERVER_HTTP_PORT: "3000"
    volumes:
      - grafana-data:/var/lib/grafana

volumes:
  prometheus-data:
  grafana-data:

Put the Grafana admin password in a .env file next to the Compose file and lock it down.

cd /opt/monitoring
printf 'GF_ADMIN_PASSWORD=%s\n' "$(openssl rand -base64 24)" | sudo tee .env >/dev/null
sudo chmod 600 .env
sudo docker compose up -d
sudo docker compose ps

Open a tunnel from your own machine and browse to Grafana:

ssh -N -L 3000:127.0.0.1:3000 -L 9090:127.0.0.1:9090 deploy@203.0.113.10

Log in at http://localhost:3000 as admin with the password from .env. Then add the data source: Connections → Data sources → Add data source → Prometheus, set the URL to http://127.0.0.1:9090, and click Save & test. You should see a green confirmation.

Import a node_exporter dashboard

You can build panels by hand, but the community dashboard "Node Exporter Full" already covers CPU, memory, disk, network and load. In Grafana go to Dashboards → New → Import, enter the dashboard ID 1860, select your Prometheus data source and import. Check the dashboard page on grafana.com for the current ID and any required exporter flags, since community dashboards change over time.

If you prefer a smaller board, create a new dashboard and add panels with queries like:

100 - (avg by (instance) (rate(node_cpu_seconds_total{mode="idle"}[5m])) * 100)

(node_filesystem_avail_bytes{mountpoint="/"} / node_filesystem_size_bytes{mountpoint="/"}) * 100

The first shows CPU usage as a percentage, the second shows free space on the root filesystem. Set the unit to percent and the graph type to Time series.

Check it works

Work through these three checks in order:

1. curl -s http://127.0.0.1:9100/metrics | wc -l returns a large number, not zero.
2. In Prometheus at http://localhost:9090, open Status → Targets. The node job should be UP with a recent scrape time.
3. In Grafana, open the imported dashboard and confirm the time range is the last 6 hours. Panels should draw lines, and CPU should move when you run sudo apt update or a similar load.

To confirm data is really flowing, run a quick CPU burn for a few seconds and watch the graph:

timeout 20 bash -c 'while :; do :; done'

Troubleshooting

• Prometheus target shows connection refused. node_exporter is stopped or listening on the wrong address. Run sudo systemctl status node_exporter and sudo ss -lntp | grep 9100. If it is bound to 127.0.0.1 and Prometheus runs in a bridge network instead of host mode, either switch to network_mode: host as shown above or bind the exporter to the Docker bridge address.
• Grafana says "No data" on every panel. The data source URL or the time range is wrong. Re-run Save & test on the Prometheus data source, then widen the dashboard time range to the last 24 hours.
• node_exporter: command not found after a reboot. The binary was copied somewhere outside PATH, or the unit file points at a different path. Check ExecStart in /etc/systemd/system/node_exporter.service and run sudo systemctl daemon-reload.
• Disk fills up quickly. Metric history is the usual cause. Lower --storage.tsdb.retention.time, raise scrape_interval from 15s to 30s, or move the prometheus-data volume to a larger disk.
• High CPU from monitoring itself. Reduce collectors with flags such as --no-collector.wifi or --no-collector.bcache, and confirm you are not scraping the same host twice from two jobs.

Next steps

• Cover the basics first with Monitor CPU, memory, disk and logs on a Linux server.
• Watch file activity alongside metrics using inotify-tools to Monitor File System Changes in Real-Time on Linux.
• Put Grafana behind HTTPS on a real domain with Nginx Proxy Manager in Docker: HTTPS for many self-hosted apps or Traefik in Docker: automatic SSL and routing by container labels.