Server monitoring should cover both infrastructure health and the user-visible application. CPU, memory and storage metrics are useful, but a server can look healthy while the website is returning errors.
What to consider
Monitor important services, response time, disk capacity, backup status and external availability. Alerts should be actionable rather than generating constant noise that the team eventually ignores.
For critical workloads, define escalation procedures and retain enough historical data to understand whether an incident was a one-time event or part of a growing capacity problem.