Closed tzumainn closed 1 month ago
MOC-R8PAC23U26
and MOC-R8PAC23U40
also have this behavior; I think I may have tested the latter with a rhelai image, but I am not positive.
@tzumainn it looks like the moc-rhelai-nvidia-1.1 image includes an EFI system partition, while most of the other images do not.
Here's how I am inspecting the images:
Download the image:
openstack image save --file moc-rhelai-nvidia-1.1.img moc-rhelai-nvidia-1.1
These are mostly qcow2 images, so we can't operate on them directly. We need to present that as block devices, which we can do by attaching them to an [nbd] device. You may first need to load the nbd
module:
modprobe nbd
Then attach the image:
qemu-nbd -c /dev/nbd0 moc-rhelai-nvidia-1.1.img
We now have a block device /dev/nbd0
and associated partition devices, like /dev/nbd0p1
.
We can inspect the partition map of these images using fdisk
:
fdisk -l /dev/nbd0
Which for the moc-rhelai-nvidia-1.1.img
image produces:
Disk /dev/nbd0: 54.54 GiB, 58566115328 bytes, 114386944 sectors
Units: sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disklabel type: gpt
Disk identifier: D209C89E-EA5E-4FBD-B161-B461CCE297E0
Device Start End Sectors Size Type
/dev/nbd0p1 2048 4095 2048 1M BIOS boot
/dev/nbd0p2 4096 1030143 1026048 501M EFI System
/dev/nbd0p3 1030144 3127295 2097152 1G Linux filesystem
/dev/nbd0p4 3127296 114386910 111259615 53.1G Linux filesystem
When you're done, detach the image:
qemu-nbd -d /dev/nbd0
Looks like we were just using images incompatible with UEFI boot, as an older centos-image worked. Closing this; I'll be opening issues for the ESI team regarding uploading new ubuntu and centos9-stream images
After provisioning this node with
moc-rhelai-nvidia-1.1
, it will no longer successfully provision any other image. The console shows messages likecouldn't find an EFI compatible partition on /dev/sda
before giving up. However, using the rhelai image still works.