VPS Health Monitoring: What to Watch and Which Tools to Use

A VPS rarely fails without warning. Disks fill up over weeks, memory leaks grow slowly, and load creeps up as traffic grows. Health monitoring catches these trends early, so you fix them on your schedule instead of during an outage. You don’t need a big monitoring stack to start: a few commands, an uptime checker, and simple alerts cover most risks.

The metrics that matter

Metric	Healthy	Warning signs
CPU usage	Varies with load	Sustained 90–100%
CPU steal (`st`)	Near 0%	Consistently above ~10%
Load average	Below the number of vCPUs	Above vCPU count for long periods
Memory	“available” well above zero	Low available + swap in heavy use; OOM kills
Disk space	< 80% used	> 90%: services start failing at 100%
Disk I/O wait (`wa`)	Low single digits	High iowait, slow queries
Network	Within plan limits	Saturated bandwidth, packet loss, unexpected spikes
Services	Running, responding	Restarts, failed units, errors in logs
Uptime from outside	200 OK, fast	Timeouts, 5xx errors

Quick health check (2 minutes)

uptime                          # load averages
free -h                         # memory and swap
df -h                           # disk space
df -i                           # inodes (can run out before space does)
vmstat 1 5                      # CPU: us, sy, id, wa, st
top -o %CPU                     # or htop: top processes
systemctl --failed              # failed services
journalctl -p err -b | tail     # errors since boot
dmesg -T | tail                 # kernel messages (OOM kills, disk errors)

See htop, the df command, and load average explained.

Monitor from outside: uptime checks

Internal metrics can’t tell you if the site is reachable for users. Add external checks:

HTTP(S) checks of your main pages (status code and response time);
TCP checks for other services (SSH, mail, game servers);
TLS certificate expiry checks.

Uptime Kuma is easy to self-host on a second small server, so it keeps working when the main one fails. See Uptime Kuma monitoring and uptime monitoring tools.

Monitoring tools by effort

Effort	Tool	What you get
Minimal	Uptime Kuma + a disk-space cron alert	Downtime and disk alerts
Low	Netdata	Real-time dashboards, hundreds of metrics, built-in alerts. See Netdata monitoring
Medium	Prometheus + node_exporter + Grafana	Long-term history, custom dashboards and alert rules. See Grafana and Prometheus monitoring
Medium	Zabbix	Agent-based monitoring with templates and SNMP for network devices. See SNMP monitoring on Linux

A simple disk-space alert

If you only add one alert, make it disk space:

#!/bin/bash
# /usr/local/bin/disk-alert.sh
THRESHOLD=85
df -P -x tmpfs -x devtmpfs | awk 'NR>1 {gsub("%","",$5); if ($5 > T) print $6, $5"%"}' T=$THRESHOLD | \
while read mount use; do
  echo "Disk $mount at $use on $(hostname)" | mail -s "Disk alert $(hostname)" [email protected]
done

sudo chmod +x /usr/local/bin/disk-alert.sh
echo "*/30 * * * * root /usr/local/bin/disk-alert.sh" | sudo tee /etc/cron.d/disk-alert

Sending mail requires a working mail setup. A webhook to Slack, Discord or Telegram with curl works just as well. See crontab guide.

Alert thresholds that work

Alert	Suggested threshold
Site down	2–3 failed checks in a row
Disk usage	85% warning, 95% critical
Memory available	< 10% for 10 minutes
CPU steal	> 15% for 15 minutes
Load	> 2 × vCPUs for 15 minutes
Certificate expiry	< 14 days
Failed systemd units	Any

Alert on things that need action. Too many noisy alerts get ignored. See alerting best practices.

Logs are health data too

Errors in application, web server and system logs often appear before outages. Rotate logs so they don’t fill the disk, and scan them for spikes in errors. See log management and logrotate.

Weekly checklist

Disk usage and growth trend
Pending security updates (apt list --upgradable)
Failed services and recurring errors in logs
Backups ran, and a test restore works occasionally
Resource trends: is RAM or CPU steadily climbing?
Certificates and domain renewals

When the metrics say “upgrade”

If CPU, RAM or disk stay near their limits after you’ve optimized (caching, tuning, cleanup), it’s time for a bigger plan. Upgrades on VPS.DO are done by opening a support ticket. See when to upgrade your VPS and VPS performance troubleshooting.

Quick reference

Check	Command
Load	`uptime`
Memory	`free -h`
Disk	`df -h`, `df -i`
CPU / steal / iowait	`vmstat 1 5`
Failed services	`systemctl --failed`
Errors	`journalctl -p err -b`

Monitoring
Linux
VPS