- Minor
Tested on 6.9.0-beta25
On a rsync operation between disk1 and a cache pool (my 3rd pool i.e. using 6.9.0 multi-pool feature), I noticed strange load on core 40 and 56 so investigate.
1. Check top 10 processes CPU usage
ps aux | sort -nrk 3,3 | head -n 10
Output (removed irrelevant entries - I checked each of the other entries and none uses core 40 and 56)
...
root 11912 15.6 0.0 0 0 ? S 12:06 34:10 [unraidd1]
root 1804 5.5 0.0 0 0 ? S 12:06 12:15 [kswapd2]
...
2. Check which core unraidd1 uses
ps -aeF | grep unraid
Output (i.e. unraidd1 uses core 56)
root 11911 2 0 0 0 36 12:06 ? 00:00:00 [unraidd0]
root 11912 2 15 0 0 56 12:06 ? 00:35:49 [unraidd1]
3. Check which core kswapd2 uses
ps -aeF | grep kswap
Output (i.e. kswapd2 uses core 40)
root 1803 2 0 0 0 32 12:06 ? 00:02:10 [kswapd0]
root 1804 2 5 0 0 40 12:06 ? 00:12:55 [kswapd2]
4. append in syslinux
append isolcpus=32-63 nohz_full=32-63 rcu_nocs=32-63 kvm_amd.avic=1 mitigations=off pcie_acs_override=downstream,multifunction
Given the name unraidd1, I'm guessing it's a process spawned by Unraid. And d1 is perhaps disk1 and the sync operation is from disk1.
So conclusion is unraid spawns a process without considering isolcpus.
kswapd0 manages swap space so I'm assuming kswapd2 is the same?
Thing is I don't use the swap memory plugin or turn on any kind of virtual memory / swap space settings so I'm a little surprised to see it putting load on the CPU + I only use 67GB out of 96GB RAM so I don't see why swap space is triggered.
The load on both processes are not that high but high enough to cause some lag while gaming.
(PS: this is unrelated to the btrfs not respecting isolcpus bug I raised previously)