License block: Vezha features
What it shows
CPU per core load in Vezha is rendered as cpu.core_pct with label core. This gives visibility not only to the average value, but also to the load profile of each core separately.
Why it matters
Average can look acceptable even when one core rests at 100% and limits service performance. A component-by-core section allows you to more accurately diagnose single-threaded limitations, failed pinning, or skew in the scheduler.
An example from practice
On a 16-core node, the overall load was around 35%, but one core was constantly hovering around 95%. After changing the worker settings, the service response stabilized without increasing the number of servers.