So I stumbled upon this bug when creating a PXE Boot Network Installer for our infrastructure on CentOS 7. The relevant boot line looks like this:<br />
<br />
KERNEL vmlinuz-centos7<br />
APPEND initrd=initrd-centos7.img inst.ks=<a href="http://xxxxx/ks-centos7.cfg">http://xxxxx/ks-centos7.cfg</a> [<a href="http://xxxxx/ks-centos7.cfg" target="_blank">^</a>] inst.sshd nameserver=x.x.x.x ip=x.x.x.x::x.x.x.1:255.255.240.0:installer:eno1:none<br />
<br />
If you incorrectly specify the interface for static configuration to one that doesn't exist, for example eth0, the boot will fail despite the fact that the correct interface was configured via DHCP and was able to download the kickstart and various files. <br />
<br />
I realize that this is probably desired behavior; however my problem with this is the error message I got:<br />
<br />
dracut-initqueue: Warning: Could not boot.<br />
<br />
With no other indication whatsoever of why it refused to boot. I finally found the fix through a good deal of trial and error. Before anyone asks why I didn't do rd.debug, I did, but it produced so much useless output that any chance of finding what went wrong was lost in the noise. If dracut is going to warn that it couldn't not boot, it needs to provide an error code or something else useful for knowing why, without turning on full-verbosity debugging.
↧