Sorry for the wall of text... This ended up a lot longer than I thought it would...
TL:DR - Looking for guide to partitioning/booting and/or help with Clover config.
Background
I recently purchased a used Dell PowerEdge R730xd to use as a home lab/self-hosting project. The intention being I would install Proxmox and play around with it and see what I wanted to add in to it later. As the server did not include any drives, I figured I would purchase a PCIe to NVMe adapter to work as the "boot" drives for the system and then fill up the 24 drive bays over time if I decided I wanted to continue with the setup.
I purchased one of the Asus Hyper M.2 x16 PCIe NVMe cards that supports up to 4 drives. To go along with it, I purchased 2x 1TB Samsung 980 Pros. I had done some research ahead of time knowing this might cause some issues, but it appeared that they could be worked through.
Installation
I installed the drives and card and turned on PCIe bifurcation for the slot. The server/iDRAC didn't see the devices, but this was expected based on prior research.
Using Dell's iDRAC, I was able to virtually attach the Proxmox .iso and boot into the installer just fine. For my Proxmox install, I chose to use "zfs (RAID1)" with both 980's as the drives. Installation appeared to go through without a problem and I rebooted to finalize the install.
At this point, the server does not recognize a boot option and hangs in the POST menu asking what to do.
Problem and Possible Solution
I was aware this might be an issue. From what I've gathered, the server won't boot because of them being NVMe in the PCIe slots. Plus the fact that they don't even appear in iDRAC or BIOs confirms this.
I had discovered this is a common issue and that people suggest using Clover as a way to "jump start" the boot process.
I found this guide where someone appears to have gone through a very similar process (although for VMware ESXi) that seemed to have enough clues to what I'd need to do.
I installed Clover to a flash drive and did the steps to move in the nvme drivers, booted into Clover, and created the "preboot.log" file. I then started to edit/create the config.plist file as they described in the guide. This is the stage where I ran into problems...
Troubleshooting and Where I Need Help
When I opened the preboot.log file and did the search for "nvme", I found multiple listings. (Copy of the preboot section below for reference.) This is where my understanding of things starts to run out and I need help.
There are 8x volumes with NVMe being referenced. (The USB listings I assume are from the Clover boot media.) Just looking at the numbers, I think this means there are 4 partitions per physical drive? I assume that the RAID1 install means things are duplicated between the 2 drives.
I did some more research and found this guide on the Proxmox forums. They mention starting into the Proxmox installer and doing a debug install to run fdisk and blkid to get the PARTUUID. The second post mentions a situation that sounded exactly like mine and provided a config file with some additional options.
I got into the debug menu and ran fdisk and blkid (results copied below). This again is where I struggle to understand what I am seeing because of my lack of understanding of file-structures/partitioning/boot records.
The Request(s)
What I was hoping to find out from this post was a few things.
- Can someone explain the different pieces of information from the fdisk and blkid commands and preboot.log? I've done some work with fixing my other Linux server in the past and remember seeing some of this, but I never fully "learned" what I was seeing. If someone has a link that explains the columns, labels, under-lying concepts, etc, that'd be great! I wasn't able to find one and I think it's because I don't know enough to even form a good query...
- Hopefully someone out there has experienced this problem and can look at what I've got and tell me what I've done wrong. I feel like I am close, but just missing/not understanding something. I fully assume I've either used the incorrect volume keys for my config, or something else in the config file. I'm leaning on the former, hence point 1.
- If anyone has a "better" way to get Proxmox to boot with my current hardware, I'd like to hear it. My plan was to get Clover working and install that on the vFlash card in the server and just have that jump start the boot on a reboot.
- Hopefully this can serve as a guide/help someone else out there.
Let me know if you need more information. I am posting this kind of late so I might not get back to your question(s) until tomorrow.
fdisk
(Please note that I had to manually type this as I only had a screenshot that I couldn't get to upload. There might be typos.)
fdisk -l
Disk /dev/nvme0n1: 932GB, 1000204886016 bytes, 1953525168 sectors
121126 cylinders, 256 heads, 63 sectors/track
Units: sectors of 1 * 512 = 512 bytes
Device Boot StartCHS EndCHS StartLBA EndLBA Sectors Size ID Type
/dev/nvme0n1p1 0,0,2 1023,255,63 1 1953525167 1953525167 931G ee EFI GPT
Disk /dev/nvme1n1: 932GB, 1000204886016 bytes, 1953525168 sectors
121126 cylinders, 256 heads, 63 sectors/track
Units: sectors of 1 * 512 = 512 bytes
Device Boot StartCHS EndCHS StartLBA EndLBA Sectors Size ID Type
/dev/nvme1n1p1 0,0,2 1023,255,63 1 1953525167 1953525167 931G ee EFI GPT
blkid
(Please note that I had to manually type this as I only had a screenshot that I couldn’t get to upload. There might be typos.)
blkid
/dev/loop1: TYPE="squashfs"
/dev/nvme0n1p3: LABEL="rpool" UUID="3906746074802172538" UUID_SUB="7826638652184430782" BLOCK_SIZE="4096" TYPE="zfs_member" PARTUUID="c182c6d2-6abb-40f7-a204-967a2b6029cc"
/dev/nvme0n1p2: UUID="63F3-E64B" BLOCK_SIZE="512" TYPE="vfat" PARTUUID="06fc76a4-ed48-4f0e-84ed-f602f5962051"
/dev/sr0: BLOCK_SIZE="2048" UUID="2023-06-22-14-56-03-00" LABEL="PVE" TYPE="iso96660" PTTYPE="PMBR"
/dev/loop0: TYPE="squashfs"
/dev/nvme1n1p2: UUID="63F6-0CF7" BLOCK_SIZE="512" TYPE="vfat" PARTUUID="8231936a-7b2c-4a96-97d6-b80393a3e7a1"
/dev/nvme1n1p3: LABEL="rpool" UUID="3906746074802172538" UUID_SUB="11940256894351019100" BLOCK_SIZE="4096" TYPE="zfs_member" PARTUUID="f57fc276-bca6-4779-a161-ebe79db3275e"
/dev/nvme0n1p1: PARTUUID="7c249bb3-b7fb-4ebf-a5ae-8d3b9b4b9ab5"
/dev/nvme1n1p1: PARTUUID="0a796a75-41a4-4f57-9c1f-97817bb30963"
preboot.log
117:268 0:000 === [ ScanVolumes ] =============================
117:268 0:000 Found 11 volumes with blockIO
117:268 0:000 - [00]: Volume: PciRoot(0x0)\Pci(0x1A,0x0)\USB(0x0,0x0)\USB(0x4,0x0)\USB(0x0,0x0)
117:273 0:005 Result of bootcode detection: bootable unknown (legacy)
117:273 0:000 - [01]: Volume: PciRoot(0x0)\Pci(0x1A,0x0)\USB(0x0,0x0)\USB(0x4,0x0)\USB(0x0,0x0)\HD(1,MBR,0x3522AA59,0x3F,0x64000)
117:276 0:003 Result of bootcode detection: bootable unknown (legacy)
117:276 0:000 label : BDU
117:276 0:000 This is SelfVolume !!
117:276 0:000 - [02]: Volume: PciRoot(0x0)\Pci(0x1A,0x0)\USB(0x0,0x0)\USB(0x4,0x0)\USB(0x0,0x0)\HD(2,MBR,0x3522AA59,0x6403F,0x70CFC1)
117:280 0:003 Result of bootcode detection: bootable unknown (legacy)
117:280 0:000 - [03]: Volume: PciRoot(0x1)\Pci(0x2,0x0)\Pci(0x0,0x0)\NVMe(0x1,BD-15-A3-31-B6-38-25-00)
117:280 0:000 Result of bootcode detection: bootable Linux (grub,linux)
117:280 0:000 - [04]: Volume: PciRoot(0x1)\Pci(0x2,0x0)\Pci(0x0,0x0)\NVMe(0x1,BD-15-A3-31-B6-38-25-00)\HD(1,GPT,7C249BB3-B7FB-4EBF-A5AE-8D3B9B4B9AB5,0x22,0x7DE)
117:280 0:000 Result of bootcode detection: bootable unknown (legacy)
117:280 0:000 - [05]: Volume: PciRoot(0x1)\Pci(0x2,0x0)\Pci(0x0,0x0)\NVMe(0x1,BD-15-A3-31-B6-38-25-00)\HD(2,GPT,06FC76A4-ED48-4F0E-84ED-F602F5962051,0x800,0x200000)
117:281 0:000 Result of bootcode detection: bootable unknown (legacy)
117:283 0:002 label : EFI
117:283 0:000 - [06]: Volume: PciRoot(0x1)\Pci(0x2,0x0)\Pci(0x0,0x0)\NVMe(0x1,BD-15-A3-31-B6-38-25-00)\HD(3,GPT,C182C6D2-6ABB-40F7-A204-967A2B6029CC,0x200800,0x7450658F)
117:283 0:000 - [07]: Volume: PciRoot(0x1)\Pci(0x2,0x1)\Pci(0x0,0x0)\NVMe(0x1,F1-1B-A3-31-B6-38-25-00)
117:283 0:000 Result of bootcode detection: bootable Linux (grub,linux)
117:283 0:000 - [08]: Volume: PciRoot(0x1)\Pci(0x2,0x1)\Pci(0x0,0x0)\NVMe(0x1,F1-1B-A3-31-B6-38-25-00)\HD(1,GPT,0A796A75-41A4-4F57-9C1F-97817BB30963,0x22,0x7DE)
117:283 0:000 Result of bootcode detection: bootable unknown (legacy)
117:283 0:000 - [09]: Volume: PciRoot(0x1)\Pci(0x2,0x1)\Pci(0x0,0x0)\NVMe(0x1,F1-1B-A3-31-B6-38-25-00)\HD(2,GPT,8231936A-7B2C-4A96-97D6-B80393A3E7A1,0x800,0x200000)
117:283 0:000 Result of bootcode detection: bootable unknown (legacy)
117:286 0:002 label : EFI
117:286 0:000 - [10]: Volume: PciRoot(0x1)\Pci(0x2,0x1)\Pci(0x0,0x0)\NVMe(0x1,F1-1B-A3-31-B6-38-25-00)\HD(3,GPT,F57FC276-BCA6-4779-A161-EBE79DB3275E,0x200800,0x7450658F)
config.plist
Boot
Timeout
5
DefaultVolume
LastBootedVolume
GUI
Custom
Entries
Path
\EFI\systemd\systemd-bootx64.efi
Title
ProxMox
Type
Linux
Volume
06FC76A4-ED48-4F0E-84ED-F602F5962051
VolumeType
Internal
Path
\EFI\systemd\systemd-bootx64.efi
Title
ProxMox
Type
Linux
Volume
8231936A-7B2C-4A96-97D6-B80393A3E7A1
VolumeType
Internal
Hey, i have had the same trouble on an DL380 G9. Those bioses don't support booting from PCIe at all. My server can't even boot from drives from the Raid controller in IT-Mode.
I would suggest, by proxmox being a hypervisor, to just install proxmox on a single SATA disk and try to boot from there. This is what I have done in the end.
You can then use your NVMes as storage pool. Also you bifurcation can always also be a problem when trying to boot from those devices.
I would also as a last call try to disable bifurcation and see if one drive will show up. Maybe then you could use 2 real PCIe slots with cheap m2 to PCIe adapters.
a ssd sata drive in a cd-drive caddy converter is a good way to get a boot drive as well
The server has 24x 2.5" bays. I have an old SSD drive that I figure I could use as a last resort to be the Proxmox boot drive and then just use the NVMe's as storage.
I was just hoping to have the Proxmox install/configuration in the NVMe RAID1 just for some minor safety in case a drive dies. From what I've read, this should be possible. I'm just lacking the knowledge to know what I've done wrong. (Mostly my lack of understanding the blkid results.)