Auto Deploy restarts PCs, but boots back into Windows

[Sebastian Roth]

@DarkSwordsman About DHCP PXE boot you want to read this: https://wiki.fogproject.org/wiki/index.php/BIOS_and_UEFI_Co-Existence

Sure some hardware just does not want to play with the exit types. But give it another try. Make sure you play with the UEFI exit types for UEFI systems.

Do you have one batch of identical client hardware or is it all different systems? Please tell us what exactly you have and there might be other users around using the same hardware and might know which setting works.

[DarkSwordsman]

@Sebastian-Roth said in Auto Deploy restarts PCs, but boots back into Windows:

https://wiki.fogproject.org/wiki/index.php/BIOS_and_UEFI_Co-Existence

We have a whole batch of identical systems. I’ll have to check when I get there what motherboard it exactly is, but I’m pretty sure they are ASUS motherboards.

I should clarify that these PCs can still PXE boot normally with Legacy + UEFI mode enabled, it’s really just the exit type that is not working. Even if there is no exit type, it can reboot after it’s done, PXE boot, and still go to the menu if there is no task and select “Boot to disk” (fog.local) by default.

If I can at least get the Exit type or, preferably, fog.local option working (so it doesn’t reimage every reboot), it would solve our issues.

[DarkSwordsman]

@Sebastian-Roth @Tom-Elliott

So I should clarify. These PCs can Legacy PXE boot just fine, and it works with FOG as expected when it boots (I.E: I can select “Deploy” from the menu and image it, as well as set a deploy or capture task and then it will automatically start imaging when I go and manually PXE boot). The real issue is that the FOG client isn’t booting into the hard disk when the default (fog.local) option is called. I haven’t been able to get it to work at all with any exit type, on both BIOS and UEFI options.

Now also to clarify, when Legacy + UEFI is selected, the traditional PXE boot as well as the UEFI Windows Boot Manager are both available, so there theoretically should be nothing stopping FOG from seeing that, at least in my eyes.

So is the issue then that the FOG client over the Legacy PXE boot is a different client that can only see Legacy boot devices (I.E: the HardDisk and not Windows Boot Manager)? Do you think that UEFI Network booting with those DHCP changes will allow the FOG client to see the disk for some reason?

I’ve dug all over the internet for FOG not booting into Windows and the best I could find were unresolved posts on this forum.

Also, here are the MoBo/BIOS specs for these PCs:

[DarkSwordsman]

So I was able to start modifying the refind_efi.conf file. I tried the following settings:

scan_delay 5
default_selection Windows
scanfor internal,external,hdbios,biosexternal

I checked and Windows is on a GPT disk, not MBR.

The current issue is that, even with the scan_delay 5, it seems like it isn’t even doing anything. I don’t even think REFIND_EFI is even being loaded, even though I see refind.efi in the fog/service/ipxe directory.

[Tom Elliott]

@DarkSwordsman I would highly recommend leaving scan_delay at 0

[DarkSwordsman]

@Tom-Elliott Okay.

I just tried booting into http://10.1.0.50/fog/service/ipxe/refind.efi from the ipxe CLI and it came to the same blinking cursor icon. This tells me that something is up with refind.efi itself, and that fog is likely actually respecting the exit type properly.

I’ll do some digging, but is there anything you can suggest for this?

[Tom Elliott]

Can you show us what you mean by that exactly?

[DarkSwordsman]

So upon further research, it seems it is picking up Windows and refind_efi is working as expected, kinda. Except, I think refind is seeing the disk that windows is on and not Windows Boot Manager itself.

@Tom-Elliott The blinking cursor is exactly as it seems, a blinking cursor in the top left of the screen, and that’s it.

It goes PXE boot > FOG Menu (selects default: Boot to Hard Disk fog.local) > screen turns black for a couple seconds > blinking cursor.

The blinking cursor error usually occurs, IIRC from experience, when you try to legacy boot right into the disk that Windows 10 is on, and not going through Windows Boot Manager.

[DarkSwordsman]

I tried to enable the graphics menu with timeout 0 and textonly 0 at least to just see if there is anything.

Refind didn’t pop up with any menu at all.

[DarkSwordsman]

So I’m kind of tired of this at the moment. I’m 90% certain it’s booting into Windows, but it’s not booting into it correctly.

I swear that in my previous IT experience, I have seen the blinking cursor issue, and it had to do with trying to legacy boot Windows instead of EFI booting through Windows Boot Manager.

If I set the boot device to UEFI Hard Disk: Windows Boot Manager it boots into Windows 10 fine, but REFIND_EFI and GRUB_FIRST_FOUND_WINDOWS both result in the blinking cursor, a common Windows issue with booting.

It also feels like refind.efi is not picking up my changes in the configuration file at all.

[DarkSwordsman]

Is there anything I can put into the fog.local parameters or Boot Options? Would there be a reason that refind.efi is not picking up my refind.conf file?

[DarkSwordsman]

Refind is definitely not picking up the config file. I restarted our VM twice and no luck.

Also, @Tom-Elliott your suggestion to cp the config file without any other parameters didn’t make much sense to me, but it apparently worked for another user. What am I missing from your cp suggestion on that thread?

Edit: Thread for reference: https://forums.fogproject.org/topic/12356/refind-conf-doesn-t-appear-to-be-used/5

[DarkSwordsman]

Currently trying to download a new version of refind, as suggested by someone on the internet somewhere.

I dunno, I’m tired and just want this to work. It’s ridiculous that it’s not working by now.

Edit: Ah, yes, it was on this site: https://forums.fogproject.org/topic/11993/refind-pxe-booting-issue

[Sebastian Roth]

@DarkSwordsman said in Auto Deploy restarts PCs, but boots back into Windows:

I just tried booting into http://10.1.0.50/fog/service/ipxe/refind.efi from the ipxe CLI and it came to the same blinking cursor icon.

To me this sounds as if this particular machine / UEFI firmware is having an issue. I don’t think it even gets as far as loading any Windows components at all!! Have you updated the UEFI firmware to the very latest version?

Maybe try booting rEFInd from a USB key just to see if that ends up in the blinking cursor as well: Format a USB stick with FAT32, create directories EFI\BOOT, download the refind.efi binary from your FOG server, put onto the USB key as EFI\BOOT\BOOTX64.EFI and add the refind.conf (you were talking about refind_efi.conf file?!?). Now try booting from this USB key and see what you get.

If that doesn’t work out you might want to try loading the 32 bit version of rEFInd (refind_ia32.efi) found in the official archive.

PS: There is no way around it. You need to stick to it and try different things to figure out why it is not working.

[Sebastian Roth]

@DarkSwordsman And here is another idea. Possibly the scan_for option if causing the hang on blinking cursor. Not sure why but possibly it is. Maybe try setting this to manual only and add a menu section at the bottom of your config - see here:

menuentry "windows" {
    volume 0:
    loader \EFI\Microsoft\Boot\bootmgfw.efi
    enabled
}

Make sure you comment all the other pre-defined menu sections and set default_selection 1 to choose this entry for you.

In that same post it’s also mentioned that rEFInd 0.11.2 (current at that time) was not working on a specific HP workstation. Reverting back to verison 0.11.0 helped. Maybe another try. You see there are lots of options and sure one will get you beyond the blinking cursor.

[DarkSwordsman]

Hey @Tom-Elliott @Sebastian-Roth I am back at the LAN center and was able to test with a new version of refind.efi from http://www.rodsbooks.com/refind/getting.html

I noticed there were two refind config files. One is refind.efi and the other is refind_linux.conf. I went into both of these and set one to 5 second timeout and the other to 10 seconds. It didn’t pick up either of these settings.

I want to understand, what do you mean by setting up the DHCP to hand out EFI binaries? I can boot into the FOG menu just fine and I can deploy images and what not. It’s just that when I select “Boot from Disk” in the FOG PXE menu, the default option, none of the bootloaders seem to work.

I can also confirm that it’s picking up the different boot loaders when I change them. Grub and Sansboot both come to the error c.main() issue or whatever, and Refind.efi comes to a blinking cursor.

So, firstly, is there something I need to do to get refind.efi to see the config files in the first place? And secondly, what do I need to set, if any, in the config files to make it see the OS?

[Sebastian Roth]

@DarkSwordsman said in Auto Deploy restarts PCs, but boots back into Windows:

One is refind.efi and the other is refind_linux.conf.

refind.efi should not be a config file. I guess you mean refind.conf here. Not sure where refind_linux.conf is coming from but that’s something we provide AFAIK! Please tell us where you got that from?!

So, firstly, is there something I need to do to get refind.efi to see the config files in the first place?

In the iPXE boot loader code generated by your FOG server the config file is being pushed to the client by name. So please open the following URL (http://x.x.x.x/fog/service/ipxe/boot.php?mac=xx:xx:xx:xx:xx:xx - use the actual FOG server IP and client MAC address instead of x’s!) in your browser and post the refind part of that here. Should look like this:

imgfetch ${boot-url}/service/ipxe/refind.conf
chain -ar ${boot-url}/service/ipxe/refind.efi

And secondly, what do I need to set, if any, in the config files to make it see the OS?

Probably best that you post the contents of your config files here as it seems they have been modified a fair bit. But feel free to take a look at the default content of refind.conf that we deliver here: https://github.com/FOGProject/fogproject/blob/master/packages/web/service/ipxe/refind.conf (latest release version 1.5.5)

[DarkSwordsman]

@Sebastian-Roth Thanks for the reply. We have to figure out the user and password for the Ubuntu server so we can edit files, but this is what I found:

In boot.php, these are the lines you mentioned:

choose --default fog.local --timeout 3000 target && goto ${target}
:fog.local
imgfetch ${boot-url}/service/ipxe/refind.conf
chain -ar ${boot-url}/service/ipxe/refind.efi || goto MENU

In our refind.conf, these are all the options (I excluded all white space):

timeout 5

textonly

scanfor internal,manual

uefi_deep_legacy_scan true

scan_delay 2

default_selection Windows

menuentry "Windows" {
    volume 0:
    ostype: "Windows"
    loader \EFI\Microsoft\Boot\bootmgfw.efi
    enabled
}

I’ll play around a bit once we get the linux credentials back. Looking forward to your reply.

[Sebastian Roth]

@DarkSwordsman Normally if you schedule a task the boot code received by the client is different than on normal boots. So when you schedule the task and then open the link as mentioned before: http://x.x.x.x/fog/service/ipxe/boot.php?mac=xx:xx:xx:xx:xx:xx - use the actual FOG server IP and client MAC address instead of x’s - you should see a very different output compared to when there is no task scheduled for this host.

We have to figure out the user and password for the Ubuntu server

https://askubuntu.com/questions/24006/how-do-i-reset-a-lost-administrative-password

[DarkSwordsman]

Hey Guys,

So I was able to experiment a little bit and I realized that, by default, it’s booting using the BIOS option, and not UEFI. I had GRUB_FIRST_HDD set for BIOS and REFIND_EFI set for UEFI, and it used GRUB.

I don’t know if that helps us identify an issue. Also, I am into computers and programming, but this whole DHCP thing and handing out efi binaries just goes over my head. Are there any guides, start to finish, on the process? I tried following the one on your wiki but I just don’t get it.