Semi-Emergency - Can't boot to CLI to swap video card
 

CentOS 7: CentOS Linux release 7.8.2003 (Core)

Current video card: NVIDIA Corporation GT218 [GeForce 405] (rev a2)
Driver: elrepo kmod-nvidia-340xx

Yes, it's old, I do not game. It's dying, the fan stops spinning and then if I do not power off, it eventually overheats and system freezes.

Ordered replacement, fanless model - MSI GeForce 1030

Using the card with dying fan, when I set the system to boot to CLI via `systemctl set-default multi-user.target` it does not boot.

Gets to where it notifies me no i8042 controller found (always does that, not an issue) and stops. I can however select the rescue boot and get to a command line.

If I can't get to the command line in normal boot though, I do not feel I can safely remove the existing archaic nvidia driver, swap cards, and install the nvidia driver suitable for this card.

Furthermore I tried swapping card just to see if new card would boot CLI normally - same issue only not even rescue mode is available.

QUESTIONS:

A) How do I figure out why it won't boot into multi-user CLI using my older card?

B) How do I make sure it will boot into rescue mode using the new card?

Thank you for suggestions.

For time being I am using old card in GUI boot mode and that works. So I can boot the system and use it, but the old card does need to go.

Using GPU in CPU not an option as it's Xeon so is no GPU in CPU.

UPDATE

tried booting to cli having removed rhgb and quiet from boot parameters.

Would not boot at all, even in rescue, had to edit during boot and add systemd.unit=graphical.target to the linux16 line to get it to boot all the way at all.

This is insane.

Try addong the kernel nomodeset option and see if it boots ok.

Success
 

Adding nomodeset to the kernel options allowed system to boot to CLI, allowing me to switch card and still boot to CLI, remove legacy nvidia drivers and install new current nvidia drivers, and it is all working again.

THANK YOU.

Human memory is not ECC
 

Since the nomodeset parameter solved my problems, I looked up what it actually does.

And that brought back memories of when I first installed the system.

I built the system using a Super Micro board, only reason I need a video card at all is so that I could use a Xeon and have ECC memory. I could not install from the USB thumb drive, but the solution I found at the time was to go from the VGA output on the motherboard to my display, had to buy a cable just for that, and do a completely CLI install.

Then while still using VGA with CLI and not using the video card, I installed the nvidia drivers, shut down, rebooted into GUI with the video card instead of VGA and it worked.

So the problem of CLI not working has always been there. Wish I had known about nomodeset parameter then, I suspect adding that as a kernel parameter when installing would have alleviated the need for that HDMI to VGA adapter cable.

What nomodeset seems to do, newer kernels will load the video drivers early even when booting to CLI to allow fancy splash screens when booting. Useless eye candy but people want it. Unfortunately not all video cards work well with it. Using nomodeset disables that, so the video drivers do not load until the GUI is starting, alleviating the problem I was experiencing.

So anyway that is why it worked and now my issues are solved.