One of my CPU cores is stuck at 100%. No Backup / Parity check running. How can I find out which service is triggering this? 

 

I have one VM running, but that is pinned to different cores. 

 

 

Any advise?

 

Edited November 22, 20214 yr by doesntaffect

Diags?

right ... I didn't think that far this morning. Thanks @JorgeB

 

Latest diags are attached., 

 

I disabled VMs, Docker, stopped the array and restarted the host.  Still one core remained at 100%.

The 100% load jumps from one core to another after each restart. 

ryzen-diagnostics-20211122-1253.zip

Dynamix file integrity plugin is running.

Removed the plugin, still core 8 at 100%. Can we dig further? Happy to help / learn how to.

Try rebooting.

Same situation, though the 100% load jumped to another core. Even before I start the array.  

don't think htop adds a lot of detail here: 

 

 

To add:

 

I changed from a Ryzen Board to Epyc a few days ago and added a RX560, Fenvi T919 and Inateck USB controller card. Non of the cards is being used atm, all VMs are off. 

Edited November 22, 20214 yr by doesntaffect

What's that running find process?

 

cachedirs?

I didnt change anything in my Plugin mix, however to narrow it down, I did following earlier:

 

• removed all the Dynamix and a few others
• Disabled VMs and Docker
• rebooted and kept the array stopped for a few mins

 

Still, one core at 100%

 

Have the array now online again, docker and VMs enabled, no VMs running. 

 

Not sure how to trail that @JorgeB

Try rebooting in safe mode.

Might be some suggestions in the attached that help, for me it manifests as shfs @ 100% when array is running, but usb kworkers are the culprit when stopped.

Possibly moving usb connections around to different (ports) controllers helped.

 

 

safemode, no gui. Array offline.

 

Same situation one core at 100%, all other cores idling. 

 

Edited November 22, 20214 yr by doesntaffect

14 minutes ago, tjb_altf4 said:

Might be some suggestions in the attached that help, for me it manifests as shfs @ 100% when array is running, but usb kworkers are the culprit when stopped.

Possibly moving usb connections around to different (ports) controllers helped.

 

 

 

That points me to the Inateck USB controller I added recently. 

CPU Govenor is set to Performance. 

I'll pull the Inateck card later and review core load.

Top shows this - 

 

Who is that "kworker/12:1+pm" guy?

 

 

Edited November 22, 20214 yr by doesntaffect

I swapped all PCIe cards in out and tried different slots. At the end I can reproduce the issue, 1 - 2 CPU cores get stuck at 100%, if I plug a PCIe card (tested with Fenvi T919, Inateck RedComets U26, Radeon RX560) into PCIe slot 6 (second from top) on the Asrock Rack RomeD8-2T board. Consequently, leaving that slot empty I can add following cards without issues:

1. RX560
2. GT710
3. Asus Hyper.2 with 4x NVMEs
4. Inateck RedComents U26
5. Fenvi T919 

without issues and CPU cores going wild.

Passthrough of GT710, Inateck and T919 to Hackintosh VM works fine now with all CPU idling (unless I stress them). 

 

Power consumption of UnRaid host with all add-on cards, idling, dropped from 103/109w to 93/103w according to UnRaid Dashboard. CPU Epyc 8-core 7232P. 

Ha, I have the same exact issue with the same exact motherboard!

I actually have an AMD 5770 gpu in slot-6, I'll move it to another one and see if the issue is gone, thanks!

Where did you place you Hyper M.2?

I placed mine in slot-1 and I'm having all kinds of errors with the drives.

For test I ran pre-clear on all 4 NVME's at the same time, two were working around 2.5GB/s while the other two were @ 250MB/s and many errors in syslog.

I have 4 Adata XPG SX8200 Pro 1TB drives, all with latest firmware.

Thank you both

 

Was actually looking for a reason I have one core a 100% on my EPYCD8-2T.  My othe machine with  RomeD8-2T deos not seem to have that issue.  Both systhem had a card in slot 6.

 

Moving my HBA off slot 6 of my EPYCD8-2T removed tha one peg core from 100%.  Weird but it "fixed" the issue.

 

 

Edited March 22, 20242 yr by AppleJon