It has come down to this. 640GB is just not enough anymore.

I have four pieces of Toshiba DT01ACA100 lying around. You may remember them from the first Red Dead Ravager build and the attempt to build the Red Death Re:Birth. I was gonna hold them for the upcoming NAS build. But then this covid 19, silicon shortage from Biden carrying forward Drumpf's plan of snubbing China and blocking countries from fabricating silicon in China, and cryptominers causing the price of CPUs and GPUs to shoot through the roof and forcing me to even scrap the Pinkeye Revenger, I decided to also mothball the NAS project. Not to mention that even 12 pieces of 1TB is a bit short for storage.

Now, here's the stupid thing. I don't know if AMD actually tests out their RAID code thoroughly, but it's broken. Then again AMD had been adamant of not letting people use RAID in Linux, the only available RAID driver for Linux is old and doesn't work for NVMe, and they refused to contact the DMRAID team to get this worked on. Online documents hints at a newer version of the drivers that support NVMe, but my attempt to get it was an utter failure, AMD said to get it from Gigabyte and Gigabyte said to get it from AMD, and after that AMD pretends that the driver doesn't exists. Frustrating.

I popped in the 4 pieces of 1TB drives and configured them for RAID 10 because 4TB is overkill for my needs and I want some form of integral stability. Started getting mysterious hanging issues and mysterious channel error codes in the kernel log.

Now I start to panic that one of the drives may have gone bad in the meantime because they were placed in a pile in the studio for too long. Quickly swapped them back because the day is over and I want dinner and sleep since it was my birthday and I had just been given a snark hunt over my expiring driving license. Great way to ruin my birthday.

The next day I have more time to check things out. I tried spreading the drives across two power cables instead of off a single cable and the issue still persists. However, upon converting the drives to RAID-0 (the previous configuration), the problem went away and suddenly there were 4 healthy 1TB drives and a storage total of 4TB.

But like I said, I don't really want 4TB. I want 2TB with some form of redundancy and integrity.

So I made the decision to leave RAID behind and use what worked for me in my last two builds: ZFS. And even better, ZFS supports SSD caching, and RAID modes that AMD does not offer. So I flipped drive mode back to SATA, removed one of the drives, replaced the removed drive with a SSD I had lying around in the studio, proceeded to boot from Ventoy into an Arch LiveCD, use Arch-chroot to access my Ubuntu root, disable the AMDRAID driver, install ZFS, rebuild the init ram disk, and create a pool that suits my need. I opted for ZFS RAID5 (gives me around 1.8TB of storage) with 88GB of SSD cache. I duly set the pool up as the /home folder. Granted, all my existing data is gone, but I can live with that.

So yeah, one door closes, another opens.