Monitoraggio VPS con Prometheus, Grafana e Alertmanager

The architecture

Prometheus collects and stores metrics (a time-series database), node_exporter exposes the VPS's system metrics (CPU, RAM, disk, network), Grafana visualizes them, Alertmanager sends the alerts. Four services, one VPS, zero external dependencies.

Prerequisites

An Ubuntu 24.04 VPS with at least 2 GB of RAM and root access. Ports 3000 (Grafana), 9090 (Prometheus) and 9093 (Alertmanager) stay on localhost or behind a reverse proxy — never expose them directly to the Internet.

Step 1 — Install Prometheus

Create the user and directories, then install the official binary (the 2.53 series is an LTS):

useradd --no-create-home --shell /usr/sbin/nologin prometheus
mkdir /etc/prometheus /var/lib/prometheus
cd /tmp
wget https://github.com/prometheus/prometheus/releases/download/v2.53.1/prometheus-2.53.1.linux-amd64.tar.gz
tar -xzf prometheus-2.53.1.linux-amd64.tar.gz
cp prometheus-2.53.1.linux-amd64/prometheus prometheus-2.53.1.linux-amd64/promtool /usr/local/bin/
chown -R prometheus:prometheus /etc/prometheus /var/lib/prometheus

Create /etc/prometheus/prometheus.yml:

global:
  scrape_interval: 15s

rule_files:
  - /etc/prometheus/rules/*.yml

alerting:
  alertmanagers:
    - static_configs:
        - targets: ['localhost:9093']

scrape_configs:
  - job_name: prometheus
    static_configs:
      - targets: ['localhost:9090']
  - job_name: node
    static_configs:
      - targets: ['localhost:9100']

Then the systemd unit /etc/systemd/system/prometheus.service:

[Unit]
Description=Prometheus
After=network.target

[Service]
User=prometheus
ExecStart=/usr/local/bin/prometheus \
  --config.file=/etc/prometheus/prometheus.yml \
  --storage.tsdb.path=/var/lib/prometheus \
  --storage.tsdb.retention.time=15d
Restart=always

[Install]
WantedBy=multi-user.target

Validate the configuration and start:

promtool check config /etc/prometheus/prometheus.yml
systemctl daemon-reload
systemctl enable --now prometheus

Step 2 — Install node_exporter

The Ubuntu package is good enough and installs the systemd service automatically:

apt install -y prometheus-node-exporter
systemctl status prometheus-node-exporter

The exporter listens on port 9100. On a powerful VPS, self-monitoring is essentially free — the scrape load is negligible on a Ryzen 9.

Step 3 — Install Grafana

Add the official repository:

install -d -m 0755 /etc/apt/keyrings
wget -q -O - https://apt.grafana.com/gpg.key | gpg --dearmor -o /etc/apt/keyrings/grafana.gpg
echo "deb [signed-by=/etc/apt/keyrings/grafana.gpg] https://apt.grafana.com stable main" > /etc/apt/sources.list.d/grafana.list
apt update && apt install -y grafana
systemctl enable --now grafana-server

Step 4 — First dashboard

Open http://VPS-IP:3000 (through an SSH tunnel ssh -L 3000:localhost:3000 if the port is filtered). Log in with admin/admin, change the password, then:

1. Connections → Data sources → Add → Prometheus, URL http://localhost:9090, Save & test. 2. Dashboards → Import → enter ID 1860 ("Node Exporter Full") → pick your Prometheus source.

You immediately get real-time CPU, RAM, disk, network and I/O graphs.

Step 5 — Install and configure Alertmanager

apt install -y prometheus-alertmanager

Edit /etc/prometheus/alertmanager.yml to route alerts to email and a Telegram bot (create the bot via @BotFather, get your chat_id from the bot's getUpdates URL):

global:
  smtp_smarthost: 'smtp.example.com:587'
  smtp_from: '[email protected]'
  smtp_auth_username: '[email protected]'
  smtp_auth_password: 'SMTPPassword'

route:
  receiver: email
  group_by: [alertname]
  routes:
    - match:
        severity: critical
      receiver: telegram

receivers:
  - name: email
    email_configs:
      - to: '[email protected]'
  - name: telegram
    telegram_configs:
      - bot_token: '123456:ABC-DEF-your-token'
        chat_id: 123456789
        message: '{{ .CommonAnnotations.summary }}'

systemctl restart prometheus-alertmanager

Step 6 — Alert rules

Create /etc/prometheus/rules/node.yml:

groups:
  - name: node
    rules:
      - alert: InstanceDown
        expr: up == 0
        for: 2m
        labels:
          severity: critical
        annotations:
          summary: "Instance {{ $labels.instance }} unreachable"
      - alert: DiskAlmostFull
        expr: (node_filesystem_avail_bytes{mountpoint="/"} / node_filesystem_size_bytes{mountpoint="/"}) * 100 < 15
        for: 10m
        labels:
          severity: warning
        annotations:
          summary: "Root filesystem over 85% full on {{ $labels.instance }}"
      - alert: HighLoad
        expr: node_load1 > 8
        for: 15m
        labels:
          severity: warning
        annotations:
          summary: "High load on {{ $labels.instance }}"

Tune the HighLoad threshold to your vCPU count. Validate and reload:

promtool check rules /etc/prometheus/rules/node.yml
systemctl reload prometheus

Verification

• curl localhost:9090/-/healthy and curl localhost:9100/metrics | head both respond.
• In Prometheus, Status → Targets: both jobs are UP.
• Alerts: your three rules show up in green.
• Real test: systemctl stop prometheus-node-exporter — the InstanceDown alert fires after 2 minutes, Telegram and email both receive it. Restart the service.

Troubleshooting

• Target DOWN: check the service is running (ss -tlnp | grep 9100) and the target really is localhost.
• Alerts not received: journalctl -u prometheus-alertmanager; for Telegram, the chat_id is numeric and the bot must have received a first message from you.
• Grafana datasource error: use http://localhost:9090, not the hostname, and make sure Prometheus is listening.

Going further

Any server running node_exporter can join the node job: add its IP to the target list in prometheus.yml, reload, and one Prometheus instance watches your whole fleet. For black-box checks — HTTP status, DNS answers, TLS certificate expiry — install the prometheus-blackbox-exporter package and wire it as an additional job; it pairs naturally with uptime alerts routed through the same Alertmanager. Dashboards for these exporters are importable from grafana.com the same way as ID 1860.