How to install Windows Server on a server when the ISO is too big?
February 16, 2021 5:27 PM   Subscribe

BLUF: How do I boot a server to OS installation media when the media source is too big to fit on a disc and the server won't boot to a USB drive?

I have perfectly legal possession of an abandoned Dell T330 server, running a malware-compromised Windows Server Essentials 2016. I have successfully transferred the ProSupport warranty into my name. My desire is to reinstall the original OS and activate it using the legitimate license key on the sticker on the back of the machine. I can get the machine to boot to a DVD. I have chatted in with Dell Support and requested a physical DVD of the installation media, but they say they cannot send me such, do not send out media, and redirect me to the Microsoft Website to download the evaluation ISO and it will activate and unlock into the correct, fully-licensed version once installed.

Here's the problem with that, and the reason why I asked Dell to send me a pressed disc to begin with: The Microsoft ISO is larger than 4.7GB and thus cannot fit on a DVD. It is 4,815,540,224 bytes, exceeding the DVD size by about 100MB (I cannot help but think this is specifically done on purpose but that's an issue for another day.).

No problem, right? Just make it into a USB key and install from that, right? This server has no option to boot to a USB key that I can find, and I call myself having looked through every menu, but it is possible I did miss it. Regardless, I am unable, as of yet, to be able to boot the thing to a USB stick.

The BIOS is fully updated. I have also tried sticking a spare SATA blu-ray reader in there but the machine will not recognize it at all in order to boot to it, else I would just burn the damn ISO to a BD-R.

Surely there must be some way I can make this happen, yes? I do not believe that this machine is simply unable to have the OS that shipped on it from the factory reinstalled on it by the end user, but my background is almost entirely with desktop hardware where I would have already had this thing up and flying months ago. I install *desktop* Windows and other OSs all day long, all the time, but I know almost nothing and have almost no experience with installing to server hardware. None of my normal tricks and tactics work on this server because it is not a desktop PC. It is Different. Because of the proclivities of the secure, restrictive server hardware and the extra 100 or so MB of data on the Microsoft ISO, I'm currently screwed.

How do I make this happen? It seems like my options are to find the secret menu to enable booting to a USB drive, or somehow doctoring the ISO image to fit on a DVD.

I've googled methods of taking excess files out of the ISO but most of the links end up on quasi-spammy pages promoting a particular product so I'm dubious but if that's what needs to happen and you can recommend a tool to use, I welcome that information.

I am not a computer scientist, a software engineer, or a enterprise administrator. I am a PC technician and hobbyist, but I am in an area I have never trod before, and I am frustrated and humbled by this situation.
posted by glonous keming to Computers & Internet (17 answers total) 2 users marked this as a favorite
Without further information, it's not clear why you're struggling to boot to USB. But another simple option would be to just burn to a double layer DVD.
posted by turkeyphant at 5:47 PM on February 16 [1 favorite]

If both the drive and your DVD burner support dual layer media, that can give you up to 8.5GB on a disc.
posted by automatronic at 5:57 PM on February 16 [1 favorite]

> it's not clear why you're struggling to boot to USB

Because the option to do so does not exist.

Other than the fact that I am incredibly stupid, I don't know how I have this giant blindspot about dual-layer DVDs but I basically have never heard of them or I thought they were some weird failed niche technology. I was not aware that it was a common capability and it's certainly one I've never used before.
posted by glonous keming at 6:08 PM on February 16 [1 favorite]

It should boot from a USB key. I have a pile of Dell server hardware of that generation and the previous one, so I have very good confidence on that. So let's focus on that path for the moment. Sorry for the vagueness, I mostly work on previous gen of hardware:
Make sure that you have the USB key inserted in to the computer
Reboot the server and watch the messages as they come up.
At some point you're going to see a message that says 'Press F11 to enter the boot menu' along with several other F-keys you can press to go into the BIOS, the Lifecycle Manager, etc.
As soon as you see that, Press 'F11'. The message will change to something like 'Entering the boot menu'.
It will still take some time to enter the boot menu.
It then takes a bit longer to scan the system for potentially bootable devices.
Eventually it will print a list of all the potentially bootable devices, and you can scroll up and down on the list, and press 'Enter' to select a bootable device.
Assuming that your USB device is actually bootable, and the port it's inserted in is enabled in the BIOS, it should show up, probably toward or at the bottom of the list.
You should then be able to select it, press enter and boot.

- You can't just copy the evaluation ISO on to the key. I know that for Windows 10 Microsoft just has you download an app which runs for a long time and then makes the key itself (instead of downloading an ISO), but I haven't fiddled with Windows Server in a couple decades, so I don't know how they do Windows Server there. The app Rufus from can handle any other OS's media, so I bet it can handle Windows Server
- It is possible to disable one or all of the USB ports on a T330 in the BIOS as I recall. If you've looked in the BIOS, that's probably in the Integrated Devices section. You want to make sure that USB 3.0 is set to Enabled and User Accessible Ports is set to All Ports On
- If you aren't comfortable with the BIOS, then try the device in one of each of the following locations, and one of them will probably be enabled:
* Front Panel
* Back panel, stack closest to the power supply
* Back panel, stack farthest from the power supply

DM if you have questions - I can't turn the results around especially promptly, but I can probably build a pretty step-by-step process for you if I must.
posted by wotsac at 6:33 PM on February 16 [3 favorites]

You may have already gotten this far, so I apologize if this is obvious, but have you actually tried looking in the bios/boot options after making and plugging in a usb stick? The behavior might be different once the drive is actually plugged in. Did you simply check for the option and stop there? (And wotsac is correct about using rufus to make such a drive and why you should try multiple ports.)

For example, my modern (2014~) BIOSes and UEFIs don't list USB booting separately. They just list USB drives as if they were normal hard drives, in that HDD priority list. Which means, of course, you have to plug them in before you power up so you can make them the default.
posted by Snijglau at 7:09 PM on February 16

This kind of bullshit is exactly why I acquired my Zalman ZM-VE300 hard disk enclosure with virtual optical drive emulation. The only machines I've ever encountered that will boot from an internal optical drive but won't also give me the option to boot from this thing when it's in virtual optical drive mode are very, very old; certainly older than 2016.

I don't believe the Zalman ZM-VE300 is sold any more but this iodd Iodd2531 appears to be the same device rebranded.
posted by flabdablet at 7:31 PM on February 16 [2 favorites]

Oh, one more thing - I haven't run in to it on the Dell hardware, but I have seen some sorts of USB keys fail to appear as bootable on some hardware. So you might try different USB keys. The Sandisks I have rattling around both worked with my R720 this week.
posted by wotsac at 7:37 PM on February 16

How to install the operating system on a Dell PowerEdge Server? (OS deployment) | Dell US

You should check out Dell's support documentation.
posted by zengargoyle at 7:41 PM on February 16 [1 favorite]

zengargoyle, how well insulated is Dell's inbuilt deployment stuff from malware running inside an installed OS? I generally try to use only installation media I'm quite sure are read-only for doing nuke-and-pave.
posted by flabdablet at 7:48 PM on February 16

zengagargoyle has the right path.

Dell's method for installing a new OS on a server is by using iDRAC. This is a miniature second server inside the box that sits on its own ethernet port in the back of the server. It does a lot of background maintenance tasks like keeping event logs and letting you observe power consumption while the main server is off doing its thing.

You use a Java utility running on another machine to connect to iDRAC over the maintenance port and upload an ISO as a 'virtual media' device. THEN you reboot the server and choose to boot from that virtual media instead of USB or DVD. That's how you can use ISOs larger than a DVD. I believe it's a pretty simple device (and the main server is shut down) when it comes to presenting an ISO image to the BIOS in the server so the chance of the ISO you downloaded yourself being compromised mid-installation is unlikely.

It's a pretty complicated thing to do from scratch, but it's well put together and documented. I bought a used R220 late last year and doing a nuke & pave took me 2-3 nights of tinkering and learning how to do all of this. An experienced person could do it in an hour I'm sure.

I also had a problem of my iDRAC being too old for the modern Java tool (The R220 has version 6), but that's fixable and memail me if you're stuck in the same situation. I think the T330 has a pretty recent one, like version 8.
posted by JoeZydeco at 8:07 PM on February 16 [3 favorites]

The cheapest dual-layer DVDs got me through with customer workstations lacking iDRAC, and said customer prohibited writable USB media on premises. DVD drives made before around 2009 might not handle dual-layer.
posted by gregoreo at 12:34 AM on February 17 [1 favorite]

JoeZydeco knows more about it than I do. I investigated it a bit, but my servers were shipped out to random places to be plugged into a port on a border router. So I turned off all the iDRAC stuff.

I assume it's like a hypervisor, maybe even running Linux or somethng on some other SOC. Not sure about the dedicated Ethernet port, but possible. You could also set it up to use a VLAN. The HPC cluster people used it quite a bit.

I modified a standard iso images using 'xorriso' and burned them onto a DVD to send out with the servers for emergency rebuid purposes. But I used PXE for the initial installs.

I'd only use it in a data center where you have things like management vlans, serial console concentrators, and the like.

The doc implied that you could totally boot from USB, there's probably just some trick to figure out.
posted by zengargoyle at 9:23 AM on February 17

Other than the fact that I am incredibly stupid, I don't know how I have this giant blindspot about dual-layer DVDs but I basically have never heard of them or I thought they were some weird failed niche technology. I was not aware that it was a common capability and it's certainly one I've never used before.

Hey, you're not stupid, I had exactly the same blind spot and didn't realise they existed either - I just googled something about DVD capacities, it came up tangentially, and I realised it might be an option.

And I'm currently being paid to design computers! (which don't have DVD drives - I haven't used one in years).
posted by automatronic at 10:50 AM on February 17

I assume it's like a hypervisor, maybe even running Linux or somethng on some other SOC. Not sure about the dedicated Ethernet port, but possible.

If it's installed, it's a dedicated port. See this diagram, callout #11. There's a BIOS option to configure the IP address of the port if you're having network conflicts or need to change address ranges, which is what tripped me up for the first hour or two.

It runs some unspecified OS on its own SOC on the card. It's an option on some machines so YMMV. Some also allow access through a front-panel USB connector somehow. When it's working it also provides a KVM into the running server which can be handy.
posted by JoeZydeco at 1:29 PM on February 17

UPDATE: I bought some DVD DL blanks and was able to burn the Microsoft ISO. I threw a spare SSD inside the thing on an empty SATA port, which I then had to manually enable in the BIOS to get it to show up. Once I did that and put the freshly-minted install DVD DL into the optical drive, I was able to boot and successfully install Windows Server 2016 Essentials. Going this route it was almost exactly like doing a normal desktop install: I did not invoke iDRAC or the Lifecycle manager. During the bog-standard installation wizard it asked me for the license key, and I actually got to scratch off the scratch-off foil stuff from the sticker on the back (first time I ever got to do that!). Installation went fine and once up to the desktop I was able to see that the OS was genuine and active. I didn't need to take any additional steps to convert the evaluation into the full retail version like I'd been advised by the Dell support rep, as it was already g2g.

So, short term goal is achieved. Why the USB stuff seems not to exist for me in the BIOS and boot options, and getting into iDRAC, I shall leave those for another day. I have iLO on a found HP ProLiant and it's pretty useful so I do intend to explore iDRAC when I'm on a more solid footing.

I thank you all for your responses! 😊
posted by glonous keming at 4:51 PM on February 18 [2 favorites]

Glad you got a result. Thanks for the update.

During the bog-standard installation wizard it asked me for the license key, and I actually got to scratch off the scratch-off foil stuff from the sticker on the back (first time I ever got to do that!).

The way I generally approach doing a nuke and pave for a malware-compromised OS lets me keep the licence key it was installed with rather than relying on those stickers, which usually don't have the same product key as the actual installation did and often require phone activation besides.

First thing is to make sure that only the drive with the OS on it is connected; then disconnect any wireless adapters and pull any Ethernet plugs; then start up the compromised OS, log in, and plug in a USB stick onto which I've preloaded the installer for Magical Jelly Bean Keyfinder.

Install and run that, print or write down the keys that it finds, shut everything down, use a Linux box that's set up not to auto-mount USB sticks to erase the now potentially compromised one and I'm good to go with the nuke and pave.

As for USB sticks not being available as a boot option: most common cause for that I've seen on desktop boxes and laptops is UEFI firmware with Secure Boot turned on and/or compatibility/legacy-boot mode turned off. Changing the appropriate setting and then rebooting usually exposes the missing option. Sometimes it still won't be exposed unless there's already a bootable USB drive plugged in, and sometimes it matters which USB socket it's plugged into: many BIOSes won't boot off a USB stick plugged into an add-on card, only those plugged into a directly connected mobo USB socket.

I've also seen desktop mobos that refuse to boot from their front USB sockets, and others that won't boot except from the specific block of sockets physically closest to the AT keyboard and mouse ones on the back panel. Some laptops also won't give you a USB boot option until you turn off an option described as something like "fast boot" or "quick boot" and/or turn on one described as "legacy USB support".

Perhaps some of this might be relevant if you do get around to further investigating this issue on your Dell box. If so, would appreciate a further update.
posted by flabdablet at 5:22 PM on February 18

« Older Cat ate something cooking with onion   |   Is the floor vibrating and, if so, why? Newer »

You are not logged in, either login or create an account to post comments