Tasks no longer balanced across CPUs in unRAID 6.11?

[grants169]

Setup:

I have 12 CPUs: 0-6, and 1-7 are used for unraid.  CPUs 2-8, 3-9, 4-10, 5-11 are isolated for VM usage, on an intel 8700K

 

When using unraid 6.9, the 4 CPUs for unraid would be balanced and all were used for docker tasks (without CPU pinning).

 

Now with unraid 6.11, I have Plex, Unmanic, Zoneminder, Nextcloud, mariadb, plus more and they ALL use CPU 7.  CPUs 0, 6, and 1 sit there and do nothing (other than underlying unraid tasks).

 

I understand I can use CPU pinning, but why?  Unraid 6.9 managed putting tasks on lesser used CPUs.  But if I have to pin containers to certain CPUs, I can still be bottlenecked depending on which container needs more CPU.  So, I'm confused as to why this is happening.  Any advice or insight would be appreciated.

 

[Arbadacarba]

Is it possible you just have kworker going haywire?

 

 

[grants169]

1 hour ago, Arbadacarba said:

Is it possible you just have kworker going haywire?

 

No, it's all the container (and other) processes that overwhelmingly favor CPU 7.  In top, press "F" and select option "last used CPU."  That lets me see what processes are using CPU 7.  See the last column in the attached image..

 

[grants169]

I don't know how I didn't notice this before, but coming from 6.9, I had CPU (HT) 7 isolated as the emulator pin for my VM.  Unraid 6.11 doesn't seem to honor isolating a HT, only the core and corresponding HT.  This is different behavior in 6.11 vs 6.9.  After UNisolating CPU (HT) 7 and rebooting everything seems to work again as per balancing the load among CPUs assigned for unRAID.