26xx v2 not supported?
-
Ah. Then the question is why Iām running into a huge performance hit when running more than one cpu in a guest. Running a loop where I create and remove a alpine docker container runs fine (.7s per execution) on one cpu but can take up to 45s when adding CPUs.
-
Do you have perf mode enabled in your BIOS? It's obviously not normal.
-
Perf mode as in performance mode for the CPU? Enabled, yes. Did some pinning of 4 cpus in dom0 and I got better numbers.. I'll guess I'll have to do some digging in turning off power management as well..
It's one of those dirt cheap motherboards from China, so I'll have som guess work to do.
-
Strangest thing, I don't see these performance hits when using Proxmox.
Another problem I'm seeing is that I have a 50% chance of actually booting xcp. Every other boot I dont get to the splash screen. Probably should do a clean install of this server, but 8.X installer doesn't work either. Seen some other people having the same problem..
-
Even with the alt kernel? What other people?
-
Alt did not work either.. "Other people" include bug #283.
-
But as of now I'm treating this as a classic PEBCAK.
-
Do you have any boot trace or serial so we can understand where it stops?
-
Only got an image of the boot (pardon the dirty screen.
), I don't have a serial port on this motherboard yet.Did some testing.. Booting with nosmp allows it (the installer) to start.. So I don't know what's going on here..
-
That might be a Xen bug at the first place, with some hardware doing strange thing.
What's your hardware? Intel servers?
-
Oh it's one of those cheap-ass mobos from AliExpress.
Setting
dom0_max_vcpu=1-1seems to allow boot. Now for benchmarking with one dom0 cpu. -
I bet on a weird chip issues that confuses Xen heavily. However, Xen devs are interested to have access to this kind of "weird" hardware.
Having a serial access in the future might be interesting (to try different Xen parameters, like noreboot and so on to see what's going on)
-
Yeah, I've got a DB9 port in the mail.. So I'm off to xen.org to do some reporting later on.
-
Just did a fresh install of XCP-ng 8.1 followed by
yum updateon a Dell R720 with two Intel Xeon CPU E5-2640 v2. No problems during the install process but testing with load on the CPUs is pending. Any suggestions how to test for the OPs performance issues? The host sucessfully came every time I rebooted it. -
@ube hello! I have the same MoBo (Atermiter X79 turbo v1.3) and same problem with XCP-NG 8.2. Can't boot in regular mode, stops exactly at the same point. Mt CPU is little bit different - 2689 V1. I can boot only with
nosmpflag (ordom0_max_vcpu=1-1), so basically there is something with multiprocessing. In BIOS multithreading is enabled and virtualisation too. With some testing tool (Memtest86 or AIDA) I do see processor info with 8 cores / 16 threads, also Hiren's BootCD PE boots fine into the portable windows.
Did you solve your problem somehow?@olivierlambert hello! Can this error be related to CentOS? Is there a way to install XCP-NG on already installed CentOS? I'm just thinking of different ways of solving my issue.
Thanks!
-
No you can't "transform" CentOS to XCP-ng.
