Windows could not apply unattend settings during pass (offlineServicing)

[bluedaze]

Issue description:

During the installation of Windows, the setup was canceled and an error message was displayed stating: "Windows could not apply unattended settings during pass [offlineServicing]."

Steps to reproduce:

1. Start Windows installation setup.
2. Proceed to the stage where Windows applies unattended settings during the offlineServicing pass.
3. The error message appears, and the installation is canceled.

Expected behavior:

The installation should proceed without any errors during the offlineServicing pass.

Actual behavior:

An error message appears, and the installation process is halted.

Screenshots:

Possible Solution:

This was during a Windows 11 install without TPM. Possible solutions may involve installing on another machine with TPM.

[kroese]

Can you show me your docker compose file?

The virtual machine has a TPM emulator, so it does not matter if the host has TPM or not. Also, the unattend.xml skips the TPM check, so even without this emulator there should be no problem starting the installation.

It is possible to skip the unattended.xml by setting MANUAL=Y in the compose file. But I would really like to figure out why automatic installation is not working on your system, so please wait with that.

The XML file for Win11 is https://github.com/dockur/windows/blob/master/assets/win11x64.xml and as you can see the offlineServicing section is totally empty. So I dont know how it can fail on that section.

Do you also get this error when you choose Windows 10 instead of 11?

[Parsival1973]

I do. I tried Win11x64 than Win10x64 and the same error.

[kroese]

@Parsival1973 Can you tell me a bit more about your machine? How much RAM does it have and which CPU? Does it work when you enable the flag for manual installation (see the FAQ)?

Also, can you show the Docker log? Maybe the automatic ISO you downloaded is different from mine, and that maybe cause the .xml file to be incorrect?

[Parsival1973]

Sure. HP Dl380G7 with Dual Intel® Xeon® CPU X5680 @ 3.33GHz and 32 GiB DDR3 Single-bit ECC. I am running this on Unraid. Assuming below is what you wanted to see. Will have to try the manual flag.

93.87% done, estimate finish Wed Jan 17 14:52:48 2024 94.03% done, estimate finish Wed Jan 17 14:52:49 2024 94.20% done, estimate finish Wed Jan 17 14:52:49 2024 94.37% done, estimate finish Wed Jan 17 14:52:48 2024 94.53% done, estimate finish Wed Jan 17 14:52:48 2024 94.70% done, estimate finish Wed Jan 17 14:52:48 2024 94.87% done, estimate finish Wed Jan 17 14:52:48 2024 95.04% done, estimate finish Wed Jan 17 14:52:48 2024 95.20% done, estimate finish Wed Jan 17 14:52:48 2024 BdsDxe: loading Boot0004 "Windows Boot Manager" from HD(2,GPT,D8CC2922-0785-4D7D-A415-670DD8CE4CE6,0x96800,0x32000)/\EFI\Microsoft\Boot\bootmgfw.efi BdsDxe: starting Boot0004 "Windows Boot Manager" from HD(2,GPT,D8CC2922-0785-4D7D-A415-670DD8CE4CE6,0x96800,0x32000)/\EFI\Microsoft\Boot\bootmgfw.efi

[Parsival1973]

Not sure how to implement the manual flag when installing docker in unraid.

[kroese]

@Parsival1973 I think you can add MANUAL as a Container Variable in unRaid and set it to Y. Also be sure to fully delete the volume that is used by your current container, so that it doesnt use any leftovers of the previous installation attempts.

I am really struggling with finding the cause of this issue. That I cannot reproduce it myself makes it difficult. And its also weird, because I assume that we are both using an identical ISO files for installation. And we are both using identical machines (because its a VM). So that makes it so strange that you guys are getting this error, while other people do not, since everything should be identical between us.

The only thing that is not identical is the CPU, because KVM is exposing it directly without any emulation. And I see that this Xeon is a little bit older, and @bluedaze his CPU has no TPM, so that one is also a bit older I suppose. These older CPU dont have certain instructions (like the AVX feature for example).

So maybe this problem only occurs when you are missing one of these CPU flags like AVX. If we knew which one, we could tell QEMU to emulate that instruction. So my only hope is that the manual installation will give a bit more helpful error message or insight on what is causing this.

EDIT: I made some changes to the XML, please pull the latest version (v1.05) again and see if it helped.

[bluedaze]

I tried yesterday on Unraid on a dual Intel Xeon CPU E5-2680 v4. I got the "Windows could not apply unattended settings during pass [offlineServicing]" error. I then tried a win10 VM and got a similar error with a code this time: 0x08000010

I tried on an AMD 5950x using a compose file from within Dockge. That install worked fine and was completed.

I saw the note about the newer XML, so I tried again on the Unraid. From the log and the content of the directory, I can see it downloaded/converted the iso and started the install, but I can not see the noVNC screen; I now get Something went wrong, the connection is closed.

Using RDP, I could establish a connection to the IP:RDPport, and after entering the Docker username (empty password) as seen in the XML file, I got the "user setup" part and the Win11 screen.

At this point, I can confirm the installation worked; the new issue is with noVNC.

[kroese]

Yes I also changed the VNC web-viewer code to use nginx instead of websockify, so I might have introduced a new (unrelated) issue there.

But if I understand correctly the new XML works? I still dont understand why the old one did not work for you (while it did for me), but maybe the Docker installation of unRaid has some settings that are different from normal Docker or something like that?

In any case, lets close this issue and if needed, open a new one for the VNC problem.

[Parsival1973]

I will note that the last time I tried to install without specifying the manual option (still have to try that) that I did also have a VNC error and could not display (in firefox) but when I copied the link over to chrome it displayed no problem.

[kroese]

@Parsival1973 If you got a Javascript error (complaining about clipboardHandler / isAndroid stuff), then you need to clear the browser cache or open an incognito tab to fix it. It should not happen again, since I now disabled all caching for the NoVNC javascript libraries.

@bluedaze I think I tracked down the VNC connection problem. It only happened when you mapped port 8006 to a different port in your composefile, or had a reverse proxy in between for example. The embedded nginx did a redirect which included the local port (8006) and didnt take into account the external port might be different. I fixed this in v1.06, so please pull that new version and see if it fixes your issue.

[Parsival1973]

Was not on Android... Firefox on Windows 10. Just FYI

[kroese]

@Parsival1973 I did not say you used Android, but there is a red error box if you have this caching issue and the javascript error has the word Android in it (even if you are on Windows). Since you said that switching browsers solved it, it sounded like you ran into this problem and just wanted to mention that instead of switching browsers, you can solve that red error by cleaning your browser cache.