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
[BUG] v1.2.0 Interactive ISO Fails to Install On Some Bare-Metal Devices #4510
Comments
Same issue; Dell Optiplex 3080 I've reset bios, upgraded bios, disabled SATA/M2, every usb port. Tried the few bios options I've seen for any issue, all same exact behaviour as this. I compared a v1.1.2 iso to the v1.2 iso, and I noticed bootx64.efi is not executable in v1.2.0 but was in v1.1.2 iso. I have no idea if this can be a problem, but my uneducated efforts to try and compare differences I noticed this. As well a change form using kernel.xz and rootfs.xz to initrd. But I really have no idea if any of this matters. |
I have the exact same on intel NUC12 i5-1240p. |
Another related (duplicate) issue: #4472 |
I think I found the cause!! I noticed the search at the top of grub.cfg tho is still looking for kernel.xz!! I thought this made sense, and immediately wrote the above. But i've been retesting, the only change I really had to do was root=live:/dev/sda1 . Which makes me wonder if the search line actually is not the problem at all :) I would of thought both kernel and initrd of failed to be loaded given the search was looking for kernel.zx? My ultimate 'work around' however was just specifying the partition for the rd.live Warning I am changing from label based to hard path, I know in my case the USB is /dev/sda. But if the system had a sata it could of been /dev/sdb, etc. Do not do this unless you know as you are using a less reliable method to load the image. |
Thanks, @slackspace-io, for the take care of it. I checked the rc5 and later versions. It looks like they all use the After you change this, you can boot into the installer as usual. Thanks! |
Yes, the only change that was required was the root=live portion. It was not needed to set hd0,msdos1 on the ($root) portions. However setting these did not break anything. When I press 'e' to edit the config, I do not gain access to the 'search ' line. So was not able to change the search line itself and test. Would the incorrect 'search' line, cause the label based root=live: to not be found ? I was able to successfully install both a nuc gen 8 ,and optiplex 3080 by setting root=live:/dev/sda1 instead of the labeled based identification. |
@slackspace-io Thanks. We can reproduce it. Your workaround works quite well! Specifying the real partition name or UUID path works, just not sure why it breaks with the label. |
Update the current status. We found there are two COS_LIVE label partitions of USB sticks. And the timeout default is 3000 seconds, so we must wait 50 minutes. So that might be the root cause of this situation. Also, I tried with the original ISO (which means no repack), and it works well. NOTE: We repack the ISO because we need to support legacy BIOS bootup BTW, we also found some ISO editors (I tried |
Some things confuse me about this.
The compressed vs uncompressed kernel ought not to matter. If that was the problem then the boot would fail at the grub boot prompt and you wouldn't get to the point where the kernel is running and trying to do stuff. There's a lot I don't yet know about how this ISO is constructed so I'm sure there's something important I'm missing here. |
I was mistaken about the partition labeling. If I use I ran builds from my desktop for rc5 and the release, and have been comparing the build output; it appears that the actions at the end to create the ISO image look substantially different, though I do not yet understand these differences, or where they come from. A lot about the build process is not evident to me yet. I am however now fairly convinced that some late change in the way the ISO is created is the cause of this problem. |
@Vicente-Cheng I updated the initial description as this seems to be both reproducible with UEFI & BIOS |
Pre Ready-For-Testing Checklist
|
Automation e2e test issue: harvester/tests#937 |
Validated that this also looks good as a workaround on bare-metal as well 😄 :
|
@Vicente-Cheng based on the documentation update working for both solutions in kvm/qemu & on bare-metal (consumer-grade) in what was mentioned above, I feel comfortable closing this out. Thanks again for the doc update on the workaround 😄 ! cc: @bk201 |
So where are we at? I just flashed 1.2 from the current releases with this bug. Is 1.2.0-patch1 available? |
Hi @Roguito , the docs have been updated with: |
That's what I get for focusing on github over the documentation. Thank you so much! |
added |
Describe the bug
Interactive ISO Install Fails to install on some bare-metal devices.
Devices reported so far:
Two paths seem to take place:
Path "A":
squashfs: version 4.0 (2009/01/31) Phillip Lougher
Path "B":
console=ttyS0
, hit cntrl+x to boota start job is running for /dev/mapper/live-rw (50min)
Related to:
Resulting in:
To Reproduce
Pre-Reqs:
Steps to reproduce the behavior:
( Either exercise Path B or allow Path A to run the course)
Expected behavior
The installer to not hit:
And other moments.
And allow the user to proceed to the entry point of the first page of the interactive iso install.
Environment
NOTE:
This is not reproducible with v1.2.0-rc5.
But is reproducible with v1.2.0-rc6.
Also unsuccessful in reproducing in qemu/kvm
Additional context
Attaching some logs:
dmesg-logs.log
etc-initrd-release.log
etc-os-release.log
journalctllogs.log
rdsosreport.txt
And additional pictures:
Update
Other Update
The text was updated successfully, but these errors were encountered: