Hacker News new | past | comments | ask | show | jobs | submit login

Btrfs still highly recommends a raid1 mode for Metadata, but for data itself, the raid-5 is fine.

I somewhat recall there being a little progress on trying to fix the remaining "write hole" issues, in the past year or two. But in general, I think there's very little pressure to do so because so very many many people run raid-5 for data already & it works great. Getting Metadata off raid1 is low priority, a nice to have.




Still, even with raid1 for metadata and raid5 for data, the kernel still shouts at you about it being EXPERIMENTAL every time you mount such a filesystem. I understand that it's best to err on the side of caution, but that notice does a good job of persisting the idea that btrfs isn't ready for prime-time use.

I use btrfs on most of my Linux systems now (though only one with raid5), except for backup disks and backups volumes: those I intend to keep on ext4 indefinitely.


Raid5 works ok until you scrub. Even scrubbing one device at a time is a barrage of random reads sustained for days at a time

I’ll very happily move back from MD raid 5 when linear scrub for parity raid lands


Great point. I need to test this more & find out.

I have seen a bunch of changes in the past year go by about how scrubbing is prioritized & what capacity it is allocated. On the one hand, unsettling. But also, good to see folks are prodding at this area.




Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: