I think you’re misreading it. The olive green in the CPU chart is idle. RAM cache taking up most of system memory is also normal on most Linux systems, even on desktop. That cache is freed for applications to use as needed.
Welp, my only calculation was “64 cpu threads * 30% load -> ~19 cores busy”, I may be guilty of rounding up too much…
The RAM usage is intresting however, since the kernel seems to be caching all it can, to point ejecting uneeded data into swap in order to retain the disk cache. If more ram is reserved by running processes, the (likely pict-rs, database services) disk access times will begin to degrade.
It could also be all 64 threads being used lightly with the scheduler trying to spread the load out evenly.
Not sure what the exact situation regarding swap in that graph, but I’ve also had the kernel preemptively use swap for rarely used chunks of memory in favor of cache when running long-running processes. Its probably relatively normal.
Thats ~19 cores pegged at 100%, eating 128GiB of ram (OS disk cache included) and bleeding onto swap. 🤯
I think you’re misreading it. The olive green in the CPU chart is idle. RAM cache taking up most of system memory is also normal on most Linux systems, even on desktop. That cache is freed for applications to use as needed.
Welp, my only calculation was “64 cpu threads * 30% load -> ~19 cores busy”, I may be guilty of rounding up too much… The RAM usage is intresting however, since the kernel seems to be caching all it can, to point ejecting uneeded data into swap in order to retain the disk cache. If more ram is reserved by running processes, the (likely pict-rs, database services) disk access times will begin to degrade.
It could also be all 64 threads being used lightly with the scheduler trying to spread the load out evenly.
Not sure what the exact situation regarding swap in that graph, but I’ve also had the kernel preemptively use swap for rarely used chunks of memory in favor of cache when running long-running processes. Its probably relatively normal.