Monitoring and Alerting: How to Know When Something Breaks

The worst way to learn about a server failure is when a customer calls to tell you. Proactive monitoring is the safety net that catches problems before anyone notices. Every self-hosted setup needs at least basic monitoring and alerting.

Uptime Kuma: Simple Service Monitoring

Uptime Kuma is a lightweight, self-hosted monitoring tool that checks your services every few seconds and alerts you when something goes wrong. It supports HTTP, HTTPS, TCP, Ping, DNS, and Push monitors. When a service fails, it sends alerts through email, Telegram, Slack, Discord, SMS, or any webhook endpoint.

Uptime Kuma also generates public status pages that you can share with customers: a clean dashboard showing the current status of every service, with uptime history and incident timelines.

Grafana: Deep Infrastructure Visibility

For deeper monitoring, Grafana provides dashboards that visualize system metrics: CPU usage, memory, disk space, network traffic, and application-specific metrics. Pair it with Prometheus for metrics collection and you get a complete picture of your infrastructure health.

What to Monitor

• Uptime: Every public-facing service checked every 30 seconds
• System resources: CPU, RAM, disk usage, and network I/O
• SSL certificates: Expiry dates with alerts 30, 14, and 7 days before expiration
• Backup status: Confirmation that nightly backups completed successfully
• Failed logins: Spikes in authentication failures that could indicate an attack
• Database performance: Query times, connection counts, and replication lag

Alerting Philosophy

Alert on symptoms, not causes. A high CPU alert is noise. A service returning 502 errors is a signal. Configure alerts so that every notification requires a human response. If an alert can be safely ignored, suppress it. Too many alerts train your team to ignore them all.

VPS1 monitors every client deployment with Uptime Kuma and Grafana, with 24/7 alerting and escalation.