Its probably fair to say that CentOS/RHEL startup mechanism is, well, broken beyond repair for anything but trivial cases. Out of the box, NFS root doesn’t work, and its very … extraordinarily … hard … to make it work. iSCSI and other connection mechanisms don’t work. This has been the case since 6.0. 6.4 continues the long tradition of working for trivial cases, and not working for anything remotely more interesting.
That is, without significant surgery.
Actually, the whole “lets start a kernel, use a file system image, then mount and pivot” concept is inherently dangerous. A failure that should be recoverable in any process step will lead to unbootable systems.
I think its probably time to use a ramdisk boot, and post mount/attach everything else. Then we can massively simplify the initramfs bits. To some degree Debian/Ubuntu do some of this, but its still something of a nightmare for non-traditional cases.
Curiously, cloud and cluster systems are non-traditional. Go figure.
Gonna think on this during my vacation. Maybe even code while the family rests.