I work on an HPC where we use cobbler to provision and Ansible to finish configuring systems. We've done 100's of installs of Rocky 9.2, 9.4 and now 9.6. I've been the advocate that we try using Proxmox. My boss was doubtful about our ability to provision proxmox VM's and thought that trying was a waste of time because it would be to hard to make work. He's also of the general opinion that servers should be hand-built and configured. PLZ HELP ME PROVE HIM WRONG!
Where do I look for more clues as to what needs to be changed?
* Proxmox VE 8.4.0
* Cobbler 3.3.3 running on a Rocky 9 physical host
* I created a VM w/o an ISO and configured it in ways that googling indicated were appropriate for this situation. (See below.)
* I configured cobbler to know about this VM and to install Rocky 9.6 on it. (When that failed I also tried 9.4. No joy.)
* When I start the VM it happily gets info from DHCP, TFTP downloads an image from cobbler, loads the kernel and then attempts to boot.
* The boot messages always stop at or near "systemd Started Journal Service"
* The first errors that I note going by are perhaps clues to the issue:
"systemd-gpt-auto-generator[1670]: EFI loader partition unknown, exiting."
"systemd-gpt-auto-generator[1670]: (The boot loader did not set EFI variable LoaderDevicePartUUID"
* There are numerous errors intermingled in the normal msgs. Most of them feature "was skipped because of an unmet condition check" but I also see a number of them in the dmesg output of another, working, real compute node freshly installed with Rocky 9.6. So I expect that at least most of these are not new to this VM situation and I haven't noticed them before because I haven't had to look.
My VM's definition:
1 socket, 2 cores [x86-64-v2-AES]
4GB RAM, BIOS: OVMF, VirtIO RNG: source=/dev/urandom
Boot order: net0, scsi0, ide2
Network Device (net0): started with VirtIO, also tried e1000e. Tried w/ and w/o firewall
Disk: jadevms01:vm-100-disk-0,iothread=1,size=32G
EFI Disk: jadevms01:vm-100-disk-1,efitype=4m,size=4M. Deleting this did not change outcome. The EFI loader msgs (above) still appear.
I unchecked the box Pre-Enroll keys b/c leaving it enabled caused the VM start to fail very early.
On a working physical compute node running Rocky 9.6, the file system /sys/firmware/efi/efivars is 128K in size with 51K used, so the 4MB size assigned automatically for this seems sufficient.
Secure Boot: I went into the config env by pressing ESC at VM start. It appears to be off, and I could not even enable it when I tried.
I did not configure our Proxmox server so I don't know if there were optional components that were not installed.
Thank you for your advice.
Where do I look for more clues as to what needs to be changed?
* Proxmox VE 8.4.0
* Cobbler 3.3.3 running on a Rocky 9 physical host
* I created a VM w/o an ISO and configured it in ways that googling indicated were appropriate for this situation. (See below.)
* I configured cobbler to know about this VM and to install Rocky 9.6 on it. (When that failed I also tried 9.4. No joy.)
* When I start the VM it happily gets info from DHCP, TFTP downloads an image from cobbler, loads the kernel and then attempts to boot.
* The boot messages always stop at or near "systemd Started Journal Service"
* The first errors that I note going by are perhaps clues to the issue:
"systemd-gpt-auto-generator[1670]: EFI loader partition unknown, exiting."
"systemd-gpt-auto-generator[1670]: (The boot loader did not set EFI variable LoaderDevicePartUUID"
* There are numerous errors intermingled in the normal msgs. Most of them feature "was skipped because of an unmet condition check" but I also see a number of them in the dmesg output of another, working, real compute node freshly installed with Rocky 9.6. So I expect that at least most of these are not new to this VM situation and I haven't noticed them before because I haven't had to look.
My VM's definition:
1 socket, 2 cores [x86-64-v2-AES]
4GB RAM, BIOS: OVMF, VirtIO RNG: source=/dev/urandom
Boot order: net0, scsi0, ide2
Network Device (net0): started with VirtIO, also tried e1000e. Tried w/ and w/o firewall
Disk: jadevms01:vm-100-disk-0,iothread=1,size=32G
EFI Disk: jadevms01:vm-100-disk-1,efitype=4m,size=4M. Deleting this did not change outcome. The EFI loader msgs (above) still appear.
I unchecked the box Pre-Enroll keys b/c leaving it enabled caused the VM start to fail very early.
On a working physical compute node running Rocky 9.6, the file system /sys/firmware/efi/efivars is 128K in size with 51K used, so the 4MB size assigned automatically for this seems sufficient.
Secure Boot: I went into the config env by pressing ESC at VM start. It appears to be off, and I could not even enable it when I tried.
I did not configure our Proxmox server so I don't know if there were optional components that were not installed.
Thank you for your advice.
Last edited: