openSUSE / agama

A service-based Linux installer
GNU General Public License v2.0
133 stars 39 forks source link

VM stuck at reboot after installation #349

Closed drmpjz closed 2 months ago

drmpjz commented 1 year ago

Environment: openSUSE 15.4/Virtualbox 6.1.38 host on a hpe DL380

d-intaller d-installer-live.x86_64-default.iso build date Dec, 2nd 2022 16:51

Selected 15.4 to be installed into the VM. After installation is completed reboot of the VM hangs on a black screen. Could only be resolved by terminating the VM.

After restarting the VM and selecting "Boot from hard disk" instead of "d-installer", same behaviour.

VM could only be started after ISO image of d-installer was detached from the configuration.

imobachgs commented 1 year ago

Hi!

I tried d-installer-live.x86_64-0.6-default-Build1.1.iso, and selecting "Boot from hard disk" took me to the screen below:

failed to boot

However, if I press F12 after starting the VM and select "hard disk", it works just fine.

But what I found interesting is that I get the same result using the openSUSE Leap Live media (openSUSE-Leap-15.4-KDE-Live-x86_64-Build19.2-Media.iso). So it might be related to our media images and VirtualBox (it works just fine in QEMU/KVM), but it does not seem to be specific to D-Installer. But we need to do more testing (like installing the system using the Leap Live media).

Is this the behavior you got?

Thank you!

drmpjz commented 1 year ago

I don't think this is exactly my problem. Let me try to explain more clearly

  1. Create a new VM and define a local (empty) virtual disk
  2. Upon first start, you are asked for an installation medium -> Select the ISO of d-installer
  3. Installation of openSUSE 15.4 selected and all steps completed successfully
  4. At the end of the installation system tries to reboot -> Hanging black screen
  5. Terminate VM
  6. Start VM again, at this point the ISO of d-installer is still attached to the VM. In the startup screen select "Boot from Harddisk" instead of "Installation". Again hanging black screen.
  7. Terminate VM
  8. Before starting the VM again, detach the d-installer ISO from the VM
  9. Successful boot

I can try to collect further information if you do provide steps to do so (albeit earliest on Thursday)

imobachgs commented 1 year ago

Hi @drmpjz,

Definitely, it looks like a different problem. It would be great if you could edit the GRUB prompt and remove the quiet option so we can find out where it gets stuck.

leap-grub

Thanks in advance!

drmpjz commented 1 year ago

That did not help unfortunately.

Things I did.

1.) Reattached the d-installer iso to the VM, hang is reproducible 2.) Removed ISO, boot successful 3.) Edited /etc/defaults/grub, removed quiet, run grub2-mkconfig -o /boot/grub2/grub.cfg 4.) Reboot VM (without d-installer iso) -> More verbose boot messages as expected 5.) Reattached the d-installer iso to the VM, same hang as before (just a black screen) -> Looks to me as if the boot information on the "Hard disk" is not found

P.S. I should be able to react quicker to replies until the end of the year since I am on vacation right now :-)

ancorgs commented 9 months ago

@drmpjz Since 2022, there have been many changes in the ISO image we use to test Agama. Moreover, we are planning to release a new version next week.

Would you mind retrying with the current image or with the upcoming image next week?

drmpjz commented 9 months ago

I'll have some time off starting Dec, 11th and will retry with the latest openSUSE and VirtualBox versions then.

drmpjz commented 8 months ago

failsafe.log live.log mediacheck.log Unfortunately the situation got worse in the mean time. Test environment: openSUSE 15.5 Kernel 5.14.21-150500.55.39-default virtualbox-7.0.12-lp155.2.13.1.x86_64 virtualbox-qt-7.0.12-lp155.2.13.1.x86_64 virtualbox-host-source-7.0.12-lp155.2.13.1.noarch virtualbox-kmp-default-7.0.12_k5.14.21_150500.55.31-lp155.2.13.1.x86_64

Test VM with 2 Cores, 4 GB Memory and 80G disk agama-live.x86_64-6.0.0-openSUSE-Build2.29.iso attached as IDE DVD boot device

Behaviour agama live and mediacheck - System boots from ISO, starts an X11 Server with just an empty wall paper. No windows, menus or other ways to interact with the X11 server. Behaviour mediacheck: Boots to a root prompt. Since I don't know the root password it is difficult to proceed from there.

Just to make sure the VM infrastructure is not broken, I attached the openSUSE 15.5 installation iso to the VM instead of the agama ISO. Installation and subsequent start of the installed VM worked without any problems.

Attached are the logs of the VM startup for the three different modes of agama.

imobachgs commented 8 months ago

Hi @drmpjz,

Thanks for giving it a try.

After you get into the X session (with the wallpaper) it should be a matter of seconds that Firefox starts and you can see Agama's user interface. If it does not happen, we would like to get some logs. Could you, please, change to a console (the root password is "linux") and get some logs for us? You can use the agama logs store command to grab Agama logs. Additionally, the full output of journactl could be useful too (let's see if we have some error from firefox).

Having said that, I am curious whether you can access remotely to Agama (pointing your browser to https://<your-vm-ip>:9090/).

Thanks in advance!

drmpjz commented 8 months ago

journalctl.txt agama_logs.zip As requested. Note since Github does not support tar.bz2 files I did repackage the agama log files. Nothing is listening on port 9090 in the VM.

imobachgs commented 8 months ago

Hi @drmpjz,

Using VirtualBox

First of all, I am not a VirtualBox user, but I gave it a try and I was able to perform an installation using the latest image available in systemsmanagement:Agama:Devel (agama-live.x86_64-7.0.0-openSUSE-Build1.6.iso). However, I use Tumbleweed instead of Leap, so there might be some difference.

/usr/libexec/cockpit-desktop is missing

Having said that, I had a closer look to your logs and I found this error:

Dec 26 15:24:28 localhost.localdomain startx[1184]: /root/.icewm/startup: line 3: /usr/libexec/cockpit-desktop: No such file or directory

I do know why, but it looks like some files are missing in the image you are running. cockpit-desktop is responsible for launching firefox, so that's the reason it does not get started. Could you give it a try to the latest image in systemsmanagement:Agama:Devel, please?

If it does not work, you could try to:

Other weird things

However, I found other weird things in the logs:

Conclusion

As a first step, I would try whether you are using the right image and, then, let's try to find out why some things are missing. Perhaps there is an explanation.

drmpjz commented 8 months ago

The new image works much better for me, I do get the installation screen and was able to install Tumbleweed.

But the original problem (VM stuck at reboot after installation) is still there. With the agama ISO attached the VM will not boot from "harddisk". One additional piece of information is that the boot process complains, that no valid boot device could be found.

As soon as the agama ISO is removed, the same virtual disk boots without problems.

P.S. For me the installed Tumbleweed does not boot into a graphical desktop. Is this the expected behavior?

imobachgs commented 8 months ago

Hi @drmpjz,

I am glad to hear that it worked. You get into the same situation I described in this comment: if Grub (which is the bootloader in this medium) exits, the system fails to boot. It works just fine in QEMU, but for some reason, it does not work in VirtualBox.

You will get the same behavior if you try any other openSUSE Live media. It does not apply to the standard openSUSE installation medium because it uses Linuxrc instead of Grub. I will try to find out more details by asking our booting experts.

To boot into the graphical desktop, select one of the desktop environments (DE) in the Software section of the installer. Otherwise, none of them is installed. If you selected a DE we might need to check whether it was installed or not.

drmpjz commented 8 months ago

Ok, more fun :-) I selected a DE and the after removing the agama ISO it hangs on starting the DE. This is not agama related, reproduced with a stock Tumbleweed install. Entered a bug for this here https://bugzilla.opensuse.org/show_bug.cgi?id=1218435

Question:Virtualbox seems not to be the best supported hypervisor with SUSE. Since I really only use the VMs to test new versions before updating my physical systems, I could switch to a different hypervisor. Which one is the preferred target?

imobachgs commented 8 months ago

I do not think there is a preferred hypervisor (but I do not know). I use QEMU (with virt-manager as GUI), but some teammates use VirtualBox. I could ask them for their experience with VirtualBox.

imobachgs commented 2 months ago

This problem is not specific to Agama, so unless someone strongly opposes it, I am closing this issue.

Thanks a lot for reporting!