In response to your sig, ....

You are sooooo right.

... here is my experience. Always use software RAID.

I do, whereever I set up a Linux server. I "inherited" one of my first Linux servers, and it came with two IDE hardware (or fake, I don't remember) RAID controllers. Of course, both were incompatible with each other, and caused nothing but trouble with old cables and old disks. I switched both controllers to be plain old IDE controllers, set up a software RAID, and never returned back to hardware or fake RAID.

One other server (major brand) I managed to get my hands on to be used as a development server at my previous job could use only SCSI disks, and only as RAID. The onboard RAID controller refused all attempts to just pass each individual disk to the OS (Linux, of course). Imagine the joy when one day the f-ing onboard RAID controller died and I lost access to all data on the server. I used parts of the server from the previous paragraphs, one or two really old PCI SCSI controllers, a bunch of SCSI cables, terminators, and adapters from 50 pin SCSI to the all-in-one Power-and-SCSI connectors on the disks, and got all SCSI disks running. Somehow, Linux could read the RAID, because it followed an old industry standard for the meta data, and so I spent a long night copying system and data to that old server and made the system boot from IDE disks in a software RAID.

Windows machines are a different thing. In my mind, there shouldn't be any valuable data on a Windows machine, so it can be reinstalled from scratch. Or restored from a backup happening in the background. That takes some time and usually does not hurt, because there is always another Windows machine available.

Having some kind of RAID never sounded reasonable, because harddisks are friendly devices. They start shaking heads, ticking, and finally grinding before separating you from your data. SMART data on newer disks is usually reliable, and warns you even before the disk starts making noises. You really have a lot of time before you need to copy your data to another disk. No wonder after decades of development. As explained in [OT] Reminder: SSDs die silently and Re^2: [OT] Reminder: SSDs die silently, SSDs are faster, but evil. No warnings, SMART data is unreliable at best, just what you would expect from a much younger technology.

Working with VMs on Windows (instead of running them on a server) suddenly kills several machines (one host and several VMs) when a disk dies. Due to the size of the VM disk images, restoring a backup takes a lot of time. And the next junk PC with a running Windows usually does NOT have those VMs available. So suddenly, having an exact and current copy of the disk online in no time (i.e. a RAID-1 mirror) becomes a real time saver.

Yes, fake RAIDs suck. But if the fake RAID does its one and only job, keeping a copy of the disk available even if the disk dies, it avoids restoring from backup, and that's good enough.

In the case of my work PC, the fake RAID had way too many false alarms, and failed to properly detect a dying SSD, wrongly claiming a different SSD had problems. I replaced one fake RAID with another fake RAID. It has better software, and I don't have to dig into how Windows implements its software RAID. Getting Windows to boot after hardware changes is hard enough, I don't want a software RAID layer on top of that nightmare.

Yes, I should really get rid of Windows on physical hardware. But that's a completely different story.

And there is another completely different story why I started to run VMs on Windows.

Alexander