In RHEL 10, systemd-repart is capable of creating the root disk partition if missing. It only needs /usr and the right cmdline to mount it as /sysusr.

However, the newly created root is empty. It is up to systemd to populate it with the basic folders.

In RHEL 10, something weird happens though: A VM with a freshly created root keeps looping in the boot logs, till it starts the dracut emergency console. If nothing is done and the dracut console is exited, it keeps looping until it re-enters it.

The only solution I could find was to mount /sysroot as r/w when inside the emergency console: 

mount -o remount,rw /sysroot

If added, the system proceeds to switch root and works fine.

Steps to reproduce:

• RHEL 10 with /usr as separate partition (assume is vda3 part)
• mount.usr=/dev/vda3  in kernel cmdline
• systemd-repart is loaded in initramfs, and also /usr/sbin/mkfs.ext4. In dracut config, add
  dracutmodules+=" systemd-repart "
  install_items+=" /usr/sbin/mkfs.ext4 " (being fixed in  https://issues.redhat.com/browse/RHEL-103385)
• repart config in /usr/lib/repart.d:

echo "[Partition]
Type=root
Label=new-root
Format=ext4
MakeDirectories=/usr
SizeMinBytes=64M" > /usr/lib/repart.d/10-root.conf

• root (vda2) is deleted
• reboot

Expected result:

boot without issues

Actual result:

boot loops infinitely