How to directly add an SSD to a PowerEdge server without entering BIOS or creating a Virtual Disk?

Question

I have a Dell PowerEdge R430, booting to Debian from an SSD. In order to get the system to "see" this first SSD, I first had to use the BIOS to create a Virtual Disk representing that Physical Disk. This was frustrating - I had no need or desire for a Virtual Disk - but I was content to accept it in the interests of getting the server up and running.

Now I'm trying to add a second SSD to the server for more storage. I've physically installed the drive following the instructions here, the drive bay shows a green status light, and I've rebooted "just in case", but the drive still doesn't show up in the output of lsblk (which is how I am accustomed to identifying external hard drives to then mount with mount//etc/fstab).

I would be surprised to hear that adding a drive to a server requires creating a new Virtual Disk to represent it, since AFAICT that requires accessing the BIOS, which in turn requires attaching a monitor and keyboard to the server.

Resources I've consulted for advice:

• How Reconfigure [sic] a Virtual disk or add additional hard drives - describes how to add hard drives to a Virtual Drive; but not how to add hard drives to the server directly, bypassing the concept of Virtual Drives.
• The aforementioned hotswap instructions
• PowerEdge Tutorials: Physical Disks and RAID Controller (PERC) on Servers is an index which links to the prior two pages, among others. I note that it says "A RAID controller can prevent [sic - presumably 'present'] groups of physical disks to the operating system for which data protection schemes such as RAID 5 or RAID 10 can be defined to protect and guarantee data integrity."
• This question suggests that the choice is "all-or-nothing" - I can either have all RAID-enabled disks, or not at all. I tried pretty hard to bypass a RAID setup while setting up the server initially, so I suspect that this option is not open to me - and, in any case, I suspect this would wipe the existing server setup I have on the original disk, so this is not an attractive option (and, again, it requires digging out a keyboard and monitor to access the BIOS)
• This answer suggests that it's possible, but gives no indication of how.
• This answer suggests that a similar approach is possible - "It will not be part of any existing RAID arrays. It will not have redundancy. But the PowerEdge will be able to access it" - (though that's dealing with a different model of server, and a 3.5" bay rather than the 2.5" SSDs I am using)

Answer 1

Required Note: I work for Dell

In order to get the system to "see" this first SSD, I first had to use the BIOS to create a Virtual Disk representing that Physical Disk. This was frustrating - I had no need or desire for a Virtual Disk - but I was content to accept it in the interests of getting the server up and running.

What you're describing is the normal behavior for the vast majority (all?) RAID cards. If you don't want virtual disks you ideally don't want a server with a RAID card then as virtual disks are the entire purpose of the RAID card. As Nikita mentioned though, you can change the controller mode.

In Dell terminology it is called HBA (host bust adapter) mode. On iDRAC 8 you do that here if the card supports it:

I would be surprised to hear that adding a drive to a server requires creating a new Virtual Disk to represent it, since AFAICT that requires accessing the BIOS, which in turn requires attaching a monitor and keyboard to the server.

No, you do not have to access BIOS to configure RAID. You usually configure the RAID through the iDRAC/IPMI/ILO for Dell/SuperMicro/HPE respectively.

In large datacenters we usually do the configuration programatically through the redfish API. Dell publishes a Terraform module, Ansible modules, Python, plus a standard RESTAPI guide - all of which will let you automatically configure the RAID. Any of your large production environment leverage one of those, often through a thing called OpenManage Enterprise.

Here is some Python for example that creates a virtual disk automatically.

This question suggests that the choice is "all-or-nothing" - I can either have all RAID-enabled disks, or not at all.

Correct, this is how most RAID cards work. Mixed mode, doing HBA and RAID mode at the same time, does exist but is rare. In 2023 there really isn't a production use case I can think of. It used to be you might have something like a server for video processing where you could have RAID for all the data and direct drive access for live editing, database servers with tiered storage or cache drives, etc but the plummeting cost of NVMe drives has pretty much eliminated that use case.

A word of warning: turning on HBA mode blows away all existing data.

A bit on how RAID works

In most servers with RAID you will see that all the drives will be physically wired into a PCIe card towards the back of the server. It depends on exactly what configuration you have but with Dell there's usually what we call the backplane where all the drives plug in from the front and then two blue cables (usually SAS cables) that run from that backplane to the RAID card.

For this reason, you can't just bypass the RAID card because it is electrically between you and the drives.

Switching from RAID mode to HBA mode breaks the drives because of how data is arranged in RAID mode. A full explanation of how RAID works is a bit too long but the BLUF version is that the RAID card arranges the data in a special way for redundancy according to an algorithm. When you turn HBA mode on you disable that algorithm from running and the drives are accessed directly. Since the direct access doesn't know how to read the patterns of data that were previously there it just sees garbage it doesn't know what to do with which means the effective destruction of the data short of performing a forensic analysis to recreate the original.