Utilization per (CPU) socket/thread
robinchrist opened this issue · comments
Hey,
Maybe I've just missed it - But it doesn't seem possible to add Load Average Graph for a specific socket (in a NUMA topology)?
This would be a great addition in order to monitor scheduling issues with load not being distributed over multiple sockets properly.
@robinchrist the load average metrics (i.e. the 1, 5, 15 minute values reported by uptime(1) also) are system-wide scheduling metrics, the kernel doesn't maintain them per-CPU or per-node.