Monster VM gets corrupt disk

[larstr]

VMware ESXi, 7.0.3, 22348816

I have a nice little VM here with 3TB ram, 20TB storage and 160 vcpus. The problem is that once we assign more than 128 cpus the disk gets corrupted if we copy files into the system. With 128 cpus or less no such problems occur. 

Guest OS is Ubuntu 22.04.3 LTS

Anyone seen something like this before?

Lars

[battybishop]

In my experience this is more related to the overcommit ratio, what is the underlying CPU logical core count? Is there anything else running on the host?

[larstr]

You can't overcommit the number of cpus for a single VM on ESXi like you can on Power. Logical core count is 256 and nothing else is running on this host.

Lars

[depping]

I would suggest filing a bug Lars, will do a search internally, but not seen this before.

[larstr]

Thank you Duncan,

We have opened SR with both VMware and Dell support. Since there seems to be little progress so far I'm also asking here, hoping that we're not the first ones to encounter this problem.

Lars

[depping]

I cannot find anything useful internally so far to be honest. You may want to ask for the SR to be escalated.

[CallistoJag]

Can I just ask, is this VM running with a BIOS or EFI?

[larstr]

CallistoJag,

The maximum number of cores with BIOS seems to be 128:

"[msg.vmx.invalidConfigForLargeVM] This virtual machine cannot be powered on because the current configuration does not support 160 CPUs. Decrease the virtual CPU count to 128 or enable EFI firmware and IOMMU support."

..so efi is our only option.

Lars