20.04 fails to boot via PXE (amd64) - "hidden .disk not found, and hard to discover"

Bug #1877618 reported by Robert Maerz
24
This bug affects 5 people
Affects Status Importance Assigned to Milestone
subiquity
Invalid
Undecided
Unassigned
casper (Ubuntu)
Expired
Wishlist
Unassigned
livecd-rootfs (Ubuntu)
Invalid
Undecided
Unassigned
ubiquity (Ubuntu)
Invalid
Undecided
Unassigned

Bug Description

The problem:
The following PXE cmdline fails to boot the 20.04 Desktop iso served via NFS but works flawlessly for 19.10 and earlier:
"imgargs vmlinuz initrd=initrd root=/dev/nfs boot=casper netboot=nfs nfsroot=<server_ip>:<iso_share_path> ip=dhcp splash quiet --"

During early boot, the only error message printed is:
"Unable to find a live file system on the network" with no further help on why the failure occoured or how to resolve it.

During debugging it was also found that the above message is printed for multiple failures during boot, which does not help to differientiate what's actually failing for what reason.

How to reproduce:
1. Download the official ubuntu desktop iso
2. Mount iso as loop on /mnt/iso
3. Copy (cp -r /mnt/iso/*) iso files to nfsroot location

No, do not use '*' use '.' as in cp -r /mnt/iso/. => as per documentation at https://wiki.ubuntu.com/LiveCDNetboot

4. Copy vmlinuz and initrd from iso to tftp location
5. Boot client PC
6. Watch boot fail

The solution (case specific):
In my case the underlying problem was caused by the .disk folder not being copied by cp to the nfsroot location. Once that was resolved, the above cmd line started working again.

Suggested fixes:
1. instead of .disk, use a folder which is not hidden by default and will hence be captured by normal copy operations
2. (significantly) improve error messages when currently the "no live file systems found" message would be triggered to better guide users to possible error causes and solutions
3. update documentation to point out importance of the .disk directory for PXE (and other) ways of booting

tags: added: rls-gg-incoming
summary: - 20.04 fails to boot via PXE (amd64)
+ 20.04 fails to boot via PXE (amd64) - "hidden .disk not found, and hard
+ to discover"
Revision history for this message
Dimitri John Ledkov (xnox) wrote :

Netboot UUID checks were added in focal
See https://launchpad.net/bugs/1847076

casper (1.428) focal; urgency=medium

  * Add UUID verification to netboot mounts. LP: #1847076

Revision history for this message
Dimitri John Ledkov (xnox) wrote :

Rejecting solution 1.

Improve casper error message is accepted.

Improve documentation is also accepted.

tags: removed: rls-gg-incoming
tags: added: rls-gg-notfixing
Revision history for this message
Dimitri John Ledkov (xnox) wrote :

Documentation:

I googled for "casper nfs" and that found https://wiki.ubuntu.com/LiveCDNetboot for me

Which explicitely says to use:

> sudo cp -a /mnt/. /srv/nfs/ubuntu-7.10-desktop

that copies the hidden directories.

Which documentation did you read when preparing the nfs mount?

Changed in subiquity:
status: New → Invalid
Changed in casper (Ubuntu):
status: New → Confirmed
importance: Undecided → Wishlist
Changed in livecd-rootfs (Ubuntu):
status: New → Invalid
Changed in ubiquity (Ubuntu):
status: New → Invalid
Changed in casper (Ubuntu):
status: Confirmed → Incomplete
description: updated
Revision history for this message
Launchpad Janitor (janitor) wrote :

[Expired for casper (Ubuntu) because there has been no activity for 60 days.]

Changed in casper (Ubuntu):
status: Incomplete → Expired
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.