Come on guys, can’t we just do Btrfs RAID5/6 already?

@prologic@twtxt.net ZFS is fine but it’s out-of-tree and extremely inflexible. If Btrfs RAID5/6 was reliable it would be fantastic. Add and remove drives at will, mix different sizes. I hear it’s mostly okay as long as you mirror the metadata (RAID1), scrub frequently, and don’t hammer it with too many random reads and writes. However, there are serious performance penalties when running scrubs on the full array and random reads and writes are the entire purpose of a filesystem.

Bcachefs has similar features (but not all of them, like sending/receiving) and it doesn’t have the giant scary warnings in the documentation. I hear it’s kind of slow and it was only merged into the kernel in version 6.7. I wouldn’t really trust it with my data.

I bought a couple more hard drives recently and I’m trying to figure out how I’m going to allocate them before badblocks completes. I have a few days to decide. :)

@mckinley@twtxt.net “Warning: The RAID 5 and RAID 6 modes of Btrfs are fatally flawed, and should not be used for “anything but testing with throw-away data.” – Yikes!. Gulp.

@bender@twtxt.net Exactly. It’s just not an option with warnings like that all over the place. Some people have had success, but I’m not risking it. https://lore.kernel.org/linux-btrfs/20200627032414.GX10769@hungrycats.org/

@mckinley@twtxt.net I am curious now, though. Doesn’t Synology use RAID Btrfs? How in the world do they do it? Researching…

@bender@twtxt.net Synology uses single-volume Btrfs on software RAID, which seems to be pretty solid in my research but that’s less flexible than ZFS. https://kb.synology.com/en-us/DSM/tutorial/What_was_the_RAID_implementation_for_Btrfs_File_System_on_SynologyNAS