Normalize Load Average by CPU Count

A load average of four means something very different on a two-core system than on an eight-core system. The monitoring mistake would be to read one metric in isolation. node_load5 divided by discovered CPU count is useful because it narrows the question, and raw load has no built-in normalization, so the dashboard and alert need to relate runnable work to available cpu capacity.

In software operations this falls under capacity-normalized alerting. A dashboard becomes much more valuable when the operator knows what a rising line can prove, what it cannot prove, and which second signal should confirm the hypothesis.

My prevention rule is: Use sustained normalized load and then inspect CPU mode, I/O wait and pressure before deciding whether compute is the bottleneck. That keeps false positives lower without weakening visibility into real degradation.

The hserver source evidence is b65d5d4. Keeping that provenance matters because monitoring logic changes over time; the article should remain connected to the exact engineering decision it describes.