PM2 là gì? Tại sao dùng PM2 trong production?

PM2 là production process manager không thể thiếu cho Node.js — cluster mode tận dụng multi-core CPU, pm2 reload cho zero-downtime deploy, max_memory_restart auto-restart khi memory leak.

• PM2 là production process manager cho Node.js — giải quyết auto-restart, clustering, logging, monitoring trong một tool.
• Ecosystem config (ecosystem.config.js): module.exports = { apps: [{ name: 'api', script: 'dist/index.js', instances: 'max', exec_mode: 'cluster', env: { NODE_ENV: 'production', PORT: 3000 }, max_memory_restart: '500M', error_file: 'logs/err.log', out_file: 'logs/out.log' }] } — pm2 start ecosystem.config.js.
• Zero-downtime reload: pm2 reload api restart từng worker một (rolling restart), không như pm2 restart kill tất cả cùng lúc — dùng reload cho deploys.
• Monitoring dashboard: pm2 monit real-time CPU/RAM per process; pm2 status overview tất cả processes.
• Log management: pm2 logs api --lines 100 xem logs; pm2 flush clear logs; tích hợp với logrotate để tránh disk full.
• Startup scripts: pm2 startup generate systemd/init script để PM2 tự start khi server reboot; pm2 save lưu current process list. max_memory_restart: tự restart process khi vượt ngưỡng RAM — safety net cho memory leaks.
• Lưu ý: PM2 cluster mode dùng shared nothing — sessions/cache phải ở Redis; pm2 reload không graceful nếu app không handle SIGINT/SIGTERM.

PM2 is an essential production process manager for Node.js — cluster mode exploits multi-core CPUs, pm2 reload enables zero-downtime deploys, and max_memory_restart auto-restarts on memory leaks.

• PM2 is a production process manager for Node.js — providing auto-restart, clustering, logging, and monitoring in a single tool.
• Ecosystem config (ecosystem.config.js): module.exports = { apps: [{ name: 'api', script: 'dist/index.js', instances: 'max', exec_mode: 'cluster', env: { NODE_ENV: 'production', PORT: 3000 }, max_memory_restart: '500M', error_file: 'logs/err.log', out_file: 'logs/out.log' }] } — start with pm2 start ecosystem.config.js.
• Zero-downtime reload: pm2 reload api restarts workers one at a time (rolling restart), unlike pm2 restart which kills all at once — use reload for deploys.
• Monitoring dashboard: pm2 monit for real-time CPU/RAM per process; pm2 status for an overview of all processes.
• Log management: pm2 logs api --lines 100 to view logs; pm2 flush to clear logs; integrate with logrotate to prevent disk from filling.
• Startup scripts: pm2 startup generates a systemd/init script so PM2 auto-starts on server reboot; pm2 save saves the current process list. max_memory_restart: auto-restarts the process when it exceeds a RAM threshold — a safety net for memory leaks.

Pitfall: PM2 cluster mode uses shared-nothing — sessions/cache must be in Redis; pm2 reload is not graceful if the app doesn't handle SIGINT/SIGTERM.