Diagnose CPU Performance Problems on a Linux Server

A Linux server is responding more slowly than expected and CPU utilization looks elevated. Walk through the diagnostic steps from a high-level host view down to the specific process, thread, or code path responsible.

Constraints & Assumptions

The source references htop and the investigation sequence, not a particular incident. Do not treat high utilization as a problem by itself; tie resource consumption to the workload's latency or throughput.

Clarifying Questions

What changed, and which user-facing metric degraded? Is the load spread across all cores or concentrated on a few? Is the process CPU-bound, waiting on I/O, being throttled, or contending for resources? Is it running inside a container?

What a Strong Answer Covers

A baseline comparison, per-core, per-process, and per-thread views, the split between user, system, and wait time, sampling profiles, resource limits, and a controlled test to verify a proposed fix.

Follow-up Questions

Why can a high load average coexist with low CPU utilization? What does a single saturated thread suggest? How do container CPU quotas affect a host that appears to have idle capacity?

Overview: Diagnose Linux CPU slowness using host and thread metrics, examining scheduling, limits, profiles, and a workload-linked verification of the bottleneck.