PXEBoot Diskless/Stateless issues

General support questions
Posts: 1
Joined: 2013/01/31 01:05:48

PXEBoot Diskless/Stateless issues

Postby EVIT » 2013/01/31 01:33:09

Hi -- Long time admin but first time poster here so be gentle with me... I'm having a stroke trying to get my system to boot via a PXE initrd image to a diskless system. The objective is to install a generic CentOS 6.3 image to a ramdrive and run it from there. No data will need to be stored/kept so rebooting the system would be akin to a brand new installation. There are several sites out there that provide how-to's on PXE boot installation - most of which require an NFS or HTTP server to serve up a root FS. This doesn't, however, meet my needs and almost none of them include the ramdisk booting component.

Here's where I am:

Main system installed and working well: CentOS (2.6.32-279.19.1.el6.x86_64)
PXEboot working well with tftp-server (0.49-7.el6)
DHCP is serving the PXE boot server information and the client is getting the menus.
dracut-kernel and -network installed (004-284.el6_3.1)

Here's my PXE detail:
kernel Node/vmlinuz
append initrd=Node/initramfs.img root=/dev/ram0 ramdisk_size=524288 rw ip=dhcp

The problem I'm hitting is that the root filesystem isn't mounting properly. I get following errors:
dracut warning: No root device "block:/dev/ram0" found.
and/or (depending on PXE variables)
RAMDISK: Couldn't find valid RAM disk image starting at 0
No filesystem could mount root

This has been driving me nuts! Does anyone have any thoughts/experience with ramdisk/stateless installations that don't use NFS/http? Help?


Some of the installation reference material:
https://coda.jlab.org/wiki/index.php/Ce ... less_Setup
http://www.linuxquestions.org/questions ... er-765393/

Posts: 1
Joined: 2017/04/10 13:46:43

Re: PXEBoot Diskless/Stateless issues

Postby chris87 » 2017/04/10 15:10:51

Hi there!

Didn't want to open another thread but I'm facing the exact same issue and hopelessly stuck.

There're about > 50 servers (likely to grow big in the future) from which one is let's say a management node running CentOS7.

The goal would be something like this:
When a new node is taken into use it boots via PXE, loads a customized Linux image/root filesystem into RAM, disconnect from the management node and starts doing it's job. It might have an external storage to preserve logs and results but the point would be that it should run completely from memory. (each server have > 256 GB so that's not an issue).

As I haven't got much experience with such things I've tried to find some tutorials online and bumped into this one:

It's great but not really what I wanted as it works by mounting the root file system via NFS. To do this I need to copy the folder containing the file system 50 times on the management node for no good reason.

Unfortuntaley most of the tutorials are hopelessly outdated and I don't seem to quite get the concept how it should be done. Should I copy everything to initrd and simply not unmount at the end? Are you aware of any tutorial/article that descibes the concept how it should be done and written preferably in this century?

Thank you in advance!