If I leave VMs running and the screen saver launches, I have a better than 50% chance a run away kernel_task will be waiting for me. MacOS 10.13.6 and Fusion 11.0.2

[cothren]

I generally run 3 linux VMs in the back ground and occasionally a Windows 10 VM.  If I leave my laptop idle, so that the screen saver launches, there is a good chance (>50%) that I will come back to a kernel_task race condition.  This is when the kernel_task consumes all the remaining CPU and runs the fans full tilt.  Once this starts, the only way to stop it is to reboot.  If I suspend the screen saver, it does not happen.  If I power down my VMs, it does not happen.  I have set the sleep mode to not put the disk to sleep, but this has not effect.

Any advice would be appreciated.

[wila]

Hi,

Some things to try.

- disable host screen save (or use a very simple screen saver that does not use much CPU)

- pass the power status to the VMs on the VM Settings -> Advanced -> Pass power status to VM

- if your linux VMs have a desktop environment, disable the screen saver then and make it blank the screen instead.

Something to check:

- how much virtual CPUs does each VM have? Is any of those VMs allocating more than half of the available cores (threads do not count)?

hope this helps,

--

Wil

[cothren]

I allocate 2Gb and 1 CPU to each of the 3 linux VMs, and 3Gb and 2 CPU to the Windows VM.  The linux run together.  I shut them down before launching the Windows.  My Mac has 8 Cores and 16Gb total.

I'm trying the "Pass power status to VM" to see if that does the trick.

[wila · Accepted solution]

Hi,

The VM allocation sounds fine for your hardware.

If passing the power to the VM did not help, then try to disable 3D acceleration on all VMs and see if that helps in resolving the issue.

--

Wil

[cothren]

So far so good.  This seems to have solved the issue.  If it survives the weekend, I'll mark you answer as correct. 

[cothren]

Passing power did the trick.  All three VMs survived a three day weekend.  Something they haven't done in memory.  Thank you so much.