I actually really like ReFS if you:
- are Windows based and have an edition that supports it (minimum is Windows for Workstations, eg not home or pro)
- for storage drives (not for the OS)
- remember to format the drives with integrity streams enable " /i:enable"
- hate fdisk (as you will never need to run it again)
- works great without storage spaces (eg I use Drive Pool over the top to JBOD them into one volume)
I guess I've been running ReFS on my pools for coming up to 10? years. Apart from the memory mgt weirdness that has come and gone with various Windows Patches/Updates, I've never had an issue with performance, reliability, or resilience. I ended up on ReFS after my RAID controller and FDISK had a fight completely scrambling my FAT. Prior to that I was using MS own "Windows Home Server" that did disk pooling but then SKU went away.
These days I've two 100TB ish pools, one on my MAIN-PC running WindowsForWorkstations 11H2, that is also my MC server (as well as everything else I do day to day). 2nd setup is a WinServer2016 that does the backups of all the PC's C: Drives and has a copy of all the media on the MAIN-PC's (I use FreeFileSync to push the media selectively after first previewing the changes) over a 10Gbe fibre link (with SSD caching writes on each pool before they are rebalanced out to spinning rust).
I've pretty satisfied with this setup as I've now got running on Windows:
- ReFS formatted discs
- Pooled together as one big vol with easy add/remove of individual HDD
- VFast copying between the pools
- Real backup and preview of changes before a commit
Okay, you've got this figured out. This is cool af and really pushing Windows to it's max. Look man, I'd be tearing my hair out here, the death of small companies running microsoft exchange / domain servers internally was a massive sigh of relief on my part. I *hated it*. So it's really sweet to see someone rocking Windows servers, in their home, in a comfortable way that works for them! That's what I *wanted* Windows server to be.
I'm Windows based on the machines I actually sit at, but the majority of my infrastructure / backend including serious storage is linux.
So what you've done is gone with disks formatted as ReFS but then pooled them on your own with DrivePool? Very, very smart. I'm a big advocate for JBOD pooling in non critical situations. RAID is... not a backup, and it's less useful than it used to be beyond extremely high availability situations where downtime is unacceptable.
I'm all about JBOD and intelligent pooling here. I use MergerFS with spinning disks, very fast cache SSDs and enterprise cloud storage. It's stunningly useful.
Can you explain this "Preview Of Changes Before A Commit" ? Are we talking about when backing up between your 2 arrays?