Monitor VPS with Grafana + Prometheus Pro

Then use Certbot to add SSL:

sudo apt install certbot python3-certbot-nginx -y
sudo certbot --nginx -d grafana.yourdomain.com

Post-Installation Checklist

• Node Exporter is running and curl localhost:9100/metrics responds
• Prometheus successfully scrapes the node job (check in Status → Targets)
• Grafana login works and data source test passes
• Dashboard displays real-time data for all metrics (CPU/RAM/Disk/Network)
• Alert rules are configured and contact points receive test notifications
• Ports 9090 and 9100 are closed from public internet
• Nginx reverse proxy with SSL is working correctly

Quality VPS and Hosting for Monitoring Systems

A good monitoring system requires stable VPS with high uptime. asiagb provides Linux VPS in Thailand and Singapore with 99.9% SLA and 24/7 Thai language support. Perfect for both monitoring servers and production workloads.

Before adding new metric endpoints, check domain health first using dnsxray.com — verify SSL, DNS, Security Headers, and Uptime for free on a single page. This helps ensure your endpoint is ready before scraping.

Summary: Professional Monitoring with Grafana + Prometheus

Once you've completed all the steps, you'll have a production-grade monitoring system that:

• Collects metrics every 15 seconds with 30 days of history
• Displays real-time dashboards accessible from any device via browser
• Automatically alerts you when CPU/RAM/Disk exceed thresholds
• Maintains security with firewall + SSL + reverse proxy

Recommended next steps include adding exporters for MySQL, Nginx, or your custom applications, as well as setting up Alertmanager for more complex alert routing. This Grafana + Prometheus system is highly scalable — suitable for single VPS through clusters of dozens of machines.

Verify Grafana + Prometheus is Working Correctly

# Verify Node Exporter is sending metrics
curl -s http://localhost:9100/metrics | grep node_cpu_seconds_total | head -3

# Verify Prometheus is healthy and scrape targets succeed
curl -s http://localhost:9090/-/healthy
curl -s "http://localhost:9090/api/v1/targets" | python3 -m json.tool | grep '"health"'

# Verify Grafana is responding
curl -s http://localhost:3000/api/health | python3 -m json.tool

# View current VPS metrics via Prometheus query
curl -s "http://localhost:9090/api/v1/query?query=100-(avg(rate(node_cpu_seconds_total{mode='idle'}[5m]))*100)" \
  | python3 -m json.tool | grep value

Common Issues & Troubleshooting

Prometheus targets show "DOWN" — Node Exporter not being scraped

Node Exporter may have stopped running or your Prometheus config is incorrect.

# Check node_exporter service
sudo systemctl status node_exporter

# Verify port 9100 is open
ss -tlnp | grep 9100

# Test curl directly
curl http://localhost:9100/metrics | head -5

# Check Prometheus config for correct target
cat /etc/prometheus/prometheus.yml | grep -A5 "job_name"

Grafana cannot connect to Prometheus — "Data source connection error"

The Prometheus URL in your Grafana data source is incorrect.

# Check if Prometheus is running
sudo systemctl status prometheus

# Test the endpoint Grafana will use
curl http://localhost:9090/api/v1/query?query=up

# Correct URL in Grafana data source:
# http://localhost:9090  (if Grafana and Prometheus are on the same machine)

Dashboard shows "No data" even though Prometheus targets are UP

Time range doesn't match or query labels don't match the actual instance.

# Check in Prometheus UI first:
# Visit http://VPS_IP:9090/graph and run this query:
up{job="node"}

# If you see results but Grafana doesn't, try:
# 1. Change Time range in Grafana to "Last 1 hour"
# 2. Check $instance variable in dashboard filtering
# 3. Refresh datasource in Grafana → Connections

Alerts not sending even though metrics exceed threshold

Alert state must stay in "Firing" for the duration specified in "For" before sending. Contact point must also be configured correctly.

# Check in Grafana:
# Alerting → Alert Rules → check State column
# If State is "Pending", waiting for "For" duration to complete
# If State is "Normal", metric hasn't actually exceeded threshold

# Test contact point by clicking "Test" in:
# Alerting → Contact Points → Edit → Send test notification

Prometheus consuming too much RAM/Disk

Metric data accumulates over time. Adjust retention policy or scrape interval.

# Check Prometheus disk usage
du -sh /var/lib/prometheus/

# Reduce retention time (set in systemd service or flag)
# Edit /etc/systemd/system/prometheus.service
# Add: --storage.tsdb.retention.time=15d

# Reduce scrape interval if you don't need fine resolution
# In prometheus.yml:
global:
  scrape_interval: 60s   # change from 15s to 60s

Frequently Asked Questions

How long does Prometheus store data?

Default is 15 days. In this article, we set it to 30 days. You can adjust with the --storage.tsdb.retention.time flag or limit by disk size with --storage.tsdb.retention.size=10GB. For a 2GB RAM VPS monitoring a single machine for 30 days, expect about 1-3 GB disk usage.

How do I monitor multiple VPS machines?

Install Node Exporter on every machine you want to monitor, then add targets to prometheus.yml on your single Prometheus server. Prometheus will scrape every target every 15 seconds. You don't need to install Prometheus on every machine — just one is enough.

How are Grafana Cloud and Self-hosted different?

Grafana Cloud has a Free Tier (10,000 metrics series, 50GB logs/month) with no server maintenance required, but your data is hosted on Grafana Labs' infrastructure. Self-hosted gives you full control, data stays on your VPS, and no metric limits — ideal for production requiring data privacy or retention longer than 30 days.