Linux VDS Running Slow Despite Normal CPU? Here Is a Diagnostic Runbook

A new technical runbook published on DEV Community addresses a common but misleading scenario where a Linux virtual server appears healthy by CPU metrics yet delivers slow or intermittent performance to users. The guide argues that CPU utilization alone cannot reveal whether a system is stalling on storage I/O, memory reclaim, network issues, cgroup throttling, or hypervisor steal time. It outlines a structured, read-only-first approach to incident diagnosis, starting with capturing a precise identity bundle including timestamps, hostname, and symptom description before collecting any metrics. The runbook walks through tools such as vmstat, mpstat, pidstat, and Linux Pressure Stall Information to isolate whether bottlenecks are guest-side or originate upstream in the infrastructure. It also cautions engineers to avoid running load-generation tools like fio or iperf3 on production systems without an approved maintenance window and explicit rate limits.