Booting issues with GRUB involving RAID and HDD's that have to be off during boot
 

Hardware: DX58SO & RS2BL080

Setup: There's 3x power supplies powering the system. One to mainboard, one to normal HDD's and Fans, and one to RAID array. The RAID is a 7 stack RAID5 using SATA interface for a SAS controller (hardware, not onboard). There's 5 SATA drives connected directly to the mainboard with the onboard RAID disabled, and the RAID card set as first boot device in sequence. The RAID is setup as a single chunk with an MBR partitioning on it contaning 3 primary partitions, ntfs, ntfs, ext4. GRUB was selected to install into the MBR during instalation. It is GRUB2.

Problem: When booting up to grub with everything on, grub will dumb to rescue mode reporting inability to find by UUID. Uppon performing the command ls, only the SATA drives show up in the list and grub is not seeing the RAID array it is installed onto and booting from. It the SATA drives are disconnected (or simply turned off) during boot, GRUB will boot correctly, showing proper menu entries, at which stage I am able to turn the SATA drives back on and things will go as normal with proper mounting of all partitions/drives.

Key notes: My current OS is highly irrelevant here as this is reproducible with any OS installed, this is an issue with GRUB and not the OS.

TLDR version: GRUB2 doesn't see the hardware RAID it's installed onto if other drives are connected to onboard SATA interface.

That almost sounds like a BIOS problem.

You might check and see if there is a BIOS flag about disk identification. I have a vague memory recall (getting old) that what sometimes happens is that the SATA connections get recognized first (the raid controller takes more time as it is checking consistency and such), thus the BIOS sees the SATA disks... and because the RAID controller hasn't responded yet, assumes it doesn't exist.

When the SATA disks are disconnected/powered down, it waits for the controllers... in which case the RAID controller eventually comes ready.

Thus when grub loads, it only sees what the BIOS has identified.

The reason for checking the BIOS flags is that it is possible that the RAID connection could be set to require its presence, and cause the BIOS to wait for that to come ready.

Linux doesn't have a problem with the disks being powered up later as it does its own device discovery pass, and finds them even though the BIOS didn't.

Wouldn't that make GRUB completely inaccessible and not even able to boot to it, since it's installed on that RAID?

The entire boot process, even accessing into BIOS is held up by waiting for the RAID controller to finish loading. I can actually enter the RAID controller config before I can enter system BIOS (though, I can key in for the system BIOS first).

If it'd help, I can make a video from my phone to show the exact process/problem.

If grub is coming from the RAID I would think so.

It is definitely odd.