-
Notifications
You must be signed in to change notification settings - Fork 60
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Produce live PXE images #105
Comments
This is great, we'd love to have PXE support at @packethost and we have infrastructure to help test this out - just let me know how we can help once you get to the point of wanting to try this. A relevant KB article for reference is https://support.packet.com/kb/articles/custom-ipxe |
If somebody wants to take it over, I started coreos/coreos-assembler#261 exactly for this reason, but there are a few things to polish that I'm not doing at the moment. |
yep. i'm also looking at PXE as part of #91 |
One note to drop in re PXE, specifically iPXE - we are tracking this issue ipxe/ipxe#90 which yields intermittent failures with iPXE as a result of a bug in OCSP certificate validation. Any eyes on this especially for test and validation cycles would be helpful. |
Discussed OOB with @ajeddeloh and @jlebon. If we allow both 1) a persistent We think live PXE on FCOS should initially:
|
In the initramfs or when running? I'm assuming it's rw in the initramfs at least. |
zstd compression would make sense for the squashfs. It can compress well and it decompresses faster than xz. The Fedora kernel includes support; squashfs-tools doesn't yet.
squashfs is always r/o, so you're right that this won't work as stated. |
Disabling Zincati in live PXE images: coreos/zincati#63 |
@bgilbert - I know there has been a lot of work here. Where do we stand? |
As point of interest, we are at Packet also looking at netboot.xyz for an
Arm boot effort.
…On Wed, Oct 2, 2019 at 11:05 AM Dusty Mabe ***@***.***> wrote:
@bgilbert <https://github.com/bgilbert> - I know there has been a lot of
work here. Where do we stand?
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#105?email_source=notifications&email_token=AAAIRZYXR75DU6CJJ5GGYK3QMS2FBA5CNFSM4GO33SFKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEAFCUUQ#issuecomment-537537106>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AAAIRZYC6XTF3ELTIPAKBF3QMS2FBANCNFSM4GO33SFA>
.
|
For workaround can i get full ostree system create squashfs and boot via dracut? |
We landed |
@cgwalters can you provide some draft howto to build live pxe image with ostree? |
If you can build using COSA (see README) then |
The latest release shipped live PXE images. There are no docs yet, and #281 should be fixed, but other than that it's done. |
Yup, sounds good. Closing. |
Container Linux supports PXE-booting to a live system, which is often used to run Container Linux both on bare metal and in certain virtualization environments (e.g. VMware). Users migrating from Container Linux, as well as users who could otherwise benefit from this deployment model, will need a way to netboot a live Fedora CoreOS system.
Container Linux's PXE support has the following properties:
/usr
filesystem image (i.e., OS code). There's a dracut module that detects the squashfs and mounts it instead of a permanent disk partition./boot
filesystem.coreos-install
, either manually or via a systemd oneshot service installed via an Ignition config. A CL system installed to disk in this way is an ordinary system with no further dependence on network booting.As a starting point, I'd propose continuing essentially the same model for Fedora CoreOS. The main initramfs will not be compiled into the kernel image, but since PXE loaders (PXELINUX and iPXE) support passing multiple initramfsen to the kernel, it shouldn't be a problem to continue storing the root filesystem in a separate initramfs. (It may provide confusing UX, however.) We'll want to provide binaries and docs for PXE-booting on both BIOS and UEFI systems, including UEFI with Secure Boot enabled.
One other point: on Container Linux, PXE-booted systems are assigned the OEM ID
diskless
[sic] for some purposes. (It's not passed on the kernel command line, but is used byupdate_engine_stub
when checking into CoreUpdate.) PXE booting is orthogonal to the platform ID (e.g., it's possible to netboot a VMware VM) so we should avoid conflating them in Fedora CoreOS.The text was updated successfully, but these errors were encountered: