server-management

PM2 suits Node.js applications needing clustering and zero-downtime reloads, while systemd works for any application as a Linux-native solution. Choose based on your runtime and whether you need built-in clustering features.

Monitor four categories: availability (uptime, health checks), performance (response time, throughput), errors (rate and types), and resources (CPU, memory, disk). Tier alerts into critical, warning, and info severity levels.

Scale vertically as a quick fix for single instances, such as adding RAM for memory pressure. Scale horizontally for sustainable growth by adding instances, and use auto-scaling when traffic is variable.

Follow five steps in order: check if the process is running, check logs for errors, check resources like disk and memory, check network ports and DNS, then check dependencies such as databases and external APIs.

Avoid running services as root, ignoring log rotation, skipping monitoring, relying on manual restarts, and having no backup schedule. Instead use non-root users, structured logging, auto-restart configuration, and regular backups.