Strange issue with booting XCP-NG

tuxen

@Appollonius said in Strange issue with booting XCP-NG:

Its only when I install the GPU and dont connect it to a monitor that it will not boot properly.

Maybe because, when there's a GPU installed but no monitor attached, the motherboard POST fails at EDID probe? As stated, some boards/BIOS require an explicit configuration in order to boot without a monitor/keyboard/mouse plugged, eg.:

https://www.supermicro.com/support/faqs/faq.cfm?faq=11902

A Former User

@tuxen Thanks for your reply, unfortunately I dont have the options ACPI - Display emulation. I use a Gigabyte Aorus X570 Pro for this server.
IOMMU is enabled on the bord.
With the Gigabyte B450 AORUS M I had zero issues, with booting in 'headless'.

A Former User

Unfortunately this issue has not been fixed, I will be waiting if someone has another suggestion for this. Though I do have another question, my Xen Orchestra continually says that the time is out of sync with the XCP-NG host. I have looked into the NTP settings, and they seemed the same. Does anyone know what this could be?

olivierlambert

Double check your NTP settings, if it tells it's not sync, it's not sync.

Note: XCP-ng now uses chrony, but in XO, ntpq -p will give you NTP stats

A Former User

Okay so today I upgraded to XCP-NG 8.2.0 but unfortunately it did not solve the issue with XCP-NG not being able to boot without a display attached to the GPU. Also I am getting the EFI_MEMMAP issue and I don't know if this issue has a relation to the GPU.

If someone has suggestions which I may try, please do so and I will test it

Danp

@appollonius Are you booting on BIOS or UEFI mode? I ran across this issue earlier today that sounds exactly like what you've described.

If this is the same issue, then I wonder why the "fix" hasn't made it into xcp-ng.

olivierlambert

We include all fixes from XS tree into XCP-ng (or those not included are completely analyzed to be sure about it)

Danp

@olivierlambert Wasn't trying to imply otherwise. I was more suspicious that their "fix" never made it into the public view.

Did you happen to review the issue I linked?

stormi

According to the issue it was fixed in CH 8.0, so it was in XCP-ng 8.0 too.

Danp

@stormi Yes... I understand that is the way it is supposed to work.

A Former User

@danp Well I use UEFI as well, so we do have the same issue here :), still curious in how this can be fixed of course.. I even upgraded to the latest XCP-ng 8.2.0 RC to see if it has been fixed but no luck so far So I keep my GPU out for a while again or I gotta get a AMD Workstation GPU.

I have tried every other solution, but no luck so far. I think its either XCP-NG or The MOBO I use (Gigabyte AORUS PRO X570).

Danp

@appollonius Have you tried installing CH 8.1 or 8.2 to see if it will boot without a monitor attached?

A Former User

@danp Ehmm no, I dont have an extra host where I can test it on.. Otherwise I wouldve done that with no problem...
I use a Quadro P400 and you?

Danp

@appollonius I have a PowerEdge R620, no additional GPU

A Former User

@danp Aha then it should boot without problems though... Strange...

Danp

@appollonius I don't have this issue. Just been following along on this thread and happened to run across the Citrix issue I posted earlier.

A Former User

@danp Ahh I thought you had the same problem, but I still find it strange and I cannot find anything about it on the interwebs...

Danp

@appollonius Did you read the issue I linked above?

A Former User

@danp Yeah though it still does not have any solution... Which is sad

Danp

@appollonius Found the patch where Citrix backported the fix to XS 7.1 -- https://github.com/xenserver/xen.pg/commit/4629bd92f3d3263da9d7fe304df422e54d56ee56

@r1 @stormi Guessing that they didn't pass this upstream to the Xen folks.

0 committed to xenserver/xen.pg

CA-316240: Fix booting on some platforms without monitor attached

Some platform UEFI firmware doesn't initialize EFI_GRAPHICS_OUTPUT_PROTOCOL
instance if there is no physically attached display present. So trying
to use it will result in NULL pointer dereference even before existing
BootServices.

Signed-off-by: Igor Druzhinin <igor.druzhinin@citrix.com>