[openstack-dev] [ironic] reducing tftp usage and separating boot control from dhcp config
Jay Pipes
jaypipes at gmail.com
Thu May 9 15:13:25 UTC 2013
On 05/09/2013 09:39 AM, Jarrod B Johnson wrote:
> Hello, wanted to measure the relative interest in two related efforts I
> could work on. In general, I'm considering bringing over much of the
> capability of xCAT's deployment facility over to Ironic, but wanted to
> highlight a couple to start with.
>
> -An openstack tailored proxydhcp server. This would mean the boot
> control aspect would be cleanly split from network identity control.
> This would allow the bootstrap program to be more adaptive to UEFI and
> BIOS style boot. Given the limitations of python, I'd probably
> implement this as a moderately standalone C program (e.g. the ability to
> get at IP_PKTINFO afaict isn't cleanly possible until python 3.3). I
> can only be confident with x86, ARM I think would work with some
> different logic, POWER would not work.
> -If a system does a PXE request, then send down a second stage
> bootloader. Then that bootloader would download third stage bootloader
> (pxelinux.0/elilo/grub2/efibootmgfw.efi/pxeboot.n12/esxboot.c32/esxboot.efi)
> and would provide/download kernel/initrd/wim/multboot modules over https
> or http. This would be iPXE based (perhaps the xNBA branch that I
> established for xCAT). A patched esxboot would also be relevant. On
> EFI Linux, elilo could work with patch (as it does for xCAT), but I
> might see about grub2 depending on whether it will do something with
> Simple filesystem or not. The impetus for not elilo would be figuring
> out the most straightforward path to target-side initrd concatenation.
> I presume there is a desire to ultimately not have many copies of the
> same initrd data, but I plan to also take advantage of initrd injection
> for other features. I can do server side injection into initrds, but it
> would mean unique initrds per target.
Hi Jarrod, I have a few starter questions for you.
1) What benefit would xCAT bring to OpenStack deployments that do not
use IBM hardware?
2) How is xCAT different from IPMI + Cobbler/PXE/tftp?
3) Given that neither the above proposed solutions is Python-based, what
plans are you thinking about for how to make the eventual solution
packagable and installation via the methods that most folks use for
deployment (Chef/Puppet/etc)
Thanks!
-jay
More information about the OpenStack-dev
mailing list