XCP-NG on Minisforum MS-01

Any specific BIOS configuration settings or other recommendations for running XCP-NG v8.3.0 on a Minisforum MS-01?

Thank you!

I don’t have one that I have tested but it should work assuming the SFP+ ports are Intel based.

HP Broadcom have worked OK for me, but tearing that system down right now and probably selling/giving it away. Realtek seem to work with XCP-ng but I haven’t tested them very hard, and mine are only gigabit.

Product says x710 and i226 controllers.

I would be more concerned with the P and E cores. How well does XCP-ng work with the E cores, or do you need to turn them off?

Did a rolling pool update on my lab this morning, while I said the Broadcom cards were working, I’m seeing faster host to host VM migrations happening now that I have Intel cards installed.

There were some discussions (XCP-ng forums) of tweaks that could be done through terminal with ethtool commands that have helped some people, generally turning off some of the offloading features (transferring the load to CPU). I never got around to testing these on my “big lab” and now it doesn’t seem like I need to test them on my “mini-lab”.

From the article:

Intel engineers are now actively working on improving the Linux virtualization infrastructure for being able to convey the P/E core differences among vCPUs so that the guest VMs can better behave in such environments.

Sounds like they are not well supported yet for virtualization workloads. My understanding is you can disable the E-cores if there is an issue.

That’s what I thought, which is why I was looking at AMD 8c/16t processors when I was finding my way to a mini-lab.

And even if Intel produces some magic, will it be able to be backported to kernel 4.19 which is where XCP-ng is still operating. So far I see no motion to backport all the newer NFS options to 4.19, things link nconnect=X to use more streams per connection. VMware defaults to nconnect=4 and it is faster by a little bit on my system, I haven’t tried increasing to 8 or 16 to see what an extreme might look like.