Setting Up Server Monitoring on a Canadian VPS

The gap between a server developing a problem and that problem affecting users is the window in which monitoring can save you. CPU running at 100% for ten minutes before a service crashes, disk space filling gradually over three days before writes start failing, memory pressure causing swap to creep up before the server becomes unresponsive: these all have warning signs that monitoring catches before users feel the result.

External uptime monitoring: the simplest layer

External monitoring checks your site's availability from outside your server. UptimeRobot's free tier monitors up to 50 URLs every five minutes from multiple locations and sends email or SMS alerts when a check fails. Configure it to check the actual homepage URL rather than just a ping to the server IP: a server that responds to ICMP pings but whose web server has crashed looks up to an IP-level check but down to an HTTP check. Verify that the HTTP response code is 200 and optionally check for a specific string in the response body to confirm the page content is correct rather than a generic error page that returns 200.

System resource monitoring on the server

External monitoring tells you whether the site is up. For visibility into why it went down or whether it is about to, you need agent-based monitoring. Netdata is an open-source monitoring agent that installs in seconds, requires no configuration to start collecting metrics, and provides a real-time dashboard showing CPU usage, memory, disk I/O, network throughput, and per-process resource consumption. For persistent metrics storage and alerting, Prometheus with Grafana is the production-grade option. Prometheus scrapes metrics from exporters (node_exporter for system metrics, mysqld_exporter for MySQL, nginx or apache exporters for web server stats) and stores them in a time-series database. Grafana provides dashboards and alerting rules with integrations to PagerDuty, Slack, and email.

Alert thresholds that work in practice

These values work for most Canadian VPS running WordPress or PHP applications: CPU above 80% sustained for more than five minutes, memory above 85% used, disk space above 80% used, disk I/O wait above 20% sustained, and PHP-FPM process pool at capacity for more than two minutes. Set up alerts for specific service failures too: an alert when MySQL stops, when the web server stops, or when SSH becomes unavailable is more immediately actionable than waiting for a resource threshold breach.