Windows Server 2019 crashes on boot on DELL R720 after installing Hyper-V role

[GeneralPavlova]

Hello,

 

I am setting up a DELL R720 as a hypervisor host using Hyper-V (at least I want to)

 

I installed Windows Server 2019 Datacentre using an ISO downloaded from VLSC, (so I assume it is the latest build).

 

The R720 has been updated to have all the latest firmware, BIOS etc. When I initially installed Windows Server 2019, it went through without an issue. It was only after I installed Hyper-V feature onto the server that it has refused to boot.

I installed Windows Server on a RAID0 config I have in the system. The RAID configs are the following: The H710P Adapter controls eight SAS drives in RAID5 mode while the H710P Mini handles two SAS drives in RAID0 mode (this is the RAID controller I want to boot from, and it is this array that Windows is installed on.)

 

The symptoms:

The system initializes the processors, all of the memory, detects all the drives and initializes the firmware interfaces. The moment that it gets to the blue Windows logo with the spinning circle it immediately reboots the server and starts initializing everything again. After numerous reboots it does eventually boot into recovery mode, but can go nowhere further from here.

 

I have gone into the RAID configuration menu from the BIOS and set the boot device as the RAID controller I want, both controllers are handled by the BIOS.

I have initialized both of the RAID arrays in the RAID configuration menu and before I restarted it for the last time I had created a partition for the RAID5 array before everything stopped working.

The server had rebooted a number of times before I had installed the Hyper-V role onto it.

 

Another thing to point is that I configured the HyperV role to store any VMs on the RAID5 drives. Could this have anything to do with it crashing on boot?

 

Windows Server was installed via a UEFI USB burned using Rufus. I did not go though iDRAC or the lifecycle controller to install it.

 

Can anyone point out a possible fix, or a list of best practices for installing Windows Server on a RAID0 array on an R720 if I have to reinstall it?

 

 

Thanks in advance!

Edited September 18, 2021 by GeneralPavlova

[kmount]

Hello,

 

Similar symptoms and some ideas at https://community.spiceworks.com/topic/2178679-installing-server-2019-with-hyper-v-role-on-obr10

 

I'd strongly discourage using RAID0... Really should be RAID1 if you have a pair of boot disks to give you some tolerance of a disk failure unlike with RAID0 where either disk failing means your server won't boot.

[GeneralPavlova]

Thank you! i'll give the suggestions a go

[kmount]

Good luck, the issue looks pretty interesting please do let us know how it goes!

[GeneralPavlova]

sure thing, i'm reckoning (and please do chime in if you think i've got it wrong ) that installing Windows 2019 without UEFI (so in this case BIOS Mode) should fix it.

 

Still really puzzled as to why hyper-V has caused this type of issue.

[GeneralPavlova]

Well, all is alright in the datacentre once again...

 

The issue was because of a buggy Windows Sever 2019 STD_CORE build. Numerous other people have had the same issues on R710s, R720s and even R730s!!! (Yes I know! xD)

 

The fault was down to Microsoft, who released a patch on January 18th - 22nd of 2019. The update, labelled January 22, 2019—KB4476976 (OS Build 17763.292) fixes the issue. This can be slipstreamed, or patched on the fly using Windows Update or by manually downloading the .MSU file from the Microsoft update catalogue. Some people have still had issues with HyperV causing boot loops even after this update, so if it still happens, install the OS as BIOS instead of UEFI. This will force you to use MBR which limits partition sizes to 2TB, so partition a small amount of the VHD (120GB or so) (and assuming you have disks configured in a RAID) to MBR and partition the rest as GPT to get the rest of your RAID/total disk capacity.

 

Using BIOS + MBR rather than UEFI is supposedly janky, but in all honesty I'd prefer to go with BIOS, but it just looks like in this case I won't need to do that anymore.

 

Hope this helps anyone else!