This is more of a question than an actual issue. We have an OS image based on CentOS 7 atomic host and I was wondering if automated installation of the distro with PXE and UEFI is supposed to work. I've tried these instructions: https://wiki.fysik.dtu.dk/it/PXE_and_UEFI , but using the vmlinuz and initrd.img files from the atomic DVD (the 1902 version), and with this grub entry:
If I try that, the installer fails with a weird error: "namespace OSTree not found". As far as I can tell, this line in the anaconda Python code fails - I guess a python module missing:
gi.require_version("OSTree", "1.0")
Not sure if anaconda is part of the initrd, and simply a different initrd would make a difference here.
I'm asking because the only thing I've found online about this is this RH bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1243543 . Besides a bunch of dead links, it has a pointer to a version that supposedly works. Diffing the two anaconda versions (I have 21.48.22.147-1, the good version is supposedly 23.16-1), I see a lot of EFI-related fixes.
The goal here is to create a disk image to deploy on actual hardware - we're booting a VM and customizing it, then we can dd that on the actual hardware box. We already have this approach working for an older platform, where we use legacy BIOS.
I've managed to make this use case work, with two tweaks to the above steps:
using inst.stage2 instead of inst.repo; this works fine even with a folder that has the DVD contents, on either HTTP or NFS. It seems to be useful for loading the squashfs image, which presumably has more stuff than the minimal ramdisk
increasing the RAM to 4GB, for 2GB I would get some errors when downloading the squashfs image (one curl error and then a 'no space left on device' message later)
So this proved to be more of a kickstart / installer understanding issue.
Hi,
This is more of a question than an actual issue. We have an OS image based on CentOS 7 atomic host and I was wondering if automated installation of the distro with PXE and UEFI is supposed to work. I've tried these instructions: https://wiki.fysik.dtu.dk/it/PXE_and_UEFI , but using the vmlinuz and initrd.img files from the atomic DVD (the 1902 version), and with this grub entry:
I'm trying to install a VM like this:
sudo virt-install --hvm --connect qemu:///system --network=bridge:virbr0 --pxe --graphics spice --boot uefi --name centos7-pxetest --ram=2048 --vcpus=2 --os-type=linux --os-variant=rhel7.6 --disk path=/data/qcows/centos7-pxetest.qcow2
If I try that, the installer fails with a weird error: "namespace OSTree not found". As far as I can tell, this line in the anaconda Python code fails - I guess a python module missing:
gi.require_version("OSTree", "1.0")
Not sure if anaconda is part of the initrd, and simply a different initrd would make a difference here.
I'm asking because the only thing I've found online about this is this RH bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1243543 . Besides a bunch of dead links, it has a pointer to a version that supposedly works. Diffing the two anaconda versions (I have 21.48.22.147-1, the good version is supposedly 23.16-1), I see a lot of EFI-related fixes.
The goal here is to create a disk image to deploy on actual hardware - we're booting a VM and customizing it, then we can dd that on the actual hardware box. We already have this approach working for an older platform, where we use legacy BIOS.