Btrfs RAID5/RAID6 Support Finally Get Fixed

[phoronix]

Phoronix: Btrfs RAID5/RAID6 Support Finally Get Fixed

Over the summer we wrote about Btrfs RAID 5/6 code was found to be unsafe and likely would need a rewrite. Months later, patches have finally emerged to fix the faulty Btrfs RAID 5/6 code...

http://www.phoronix.com/scan.php?page=news_item&px=Btrfs-RAID5-RAID6-Fixed

[M@yeulC]

Nice to see this bug gone. However, RAID 5/6 is still marked as unstable on the wiki: https://btrfs.wiki.kernel.org/index.php/Status

Are there more things to address (besides code quality), or is it "production-ready"?

[89c51]

Nice to see this bug gone. However, RAID 5/6 is still marked as unstable on the wiki: https://btrfs.wiki.kernel.org/index.php/Status

Are there more things to address (besides code quality), or is it "production-ready"?

I assume it has to be merged in mainline before its marked stable in the wiki. And i hope someone is testing the hell out of it before it gets merged and someone trusts it for his data.

[Nille]

Does this need a new on disk format or its just a driver thing?

[waxhead]

Does this need a new on disk format or its just a driver thing?

From what I was able to piece together it is just a race condition and does not require a format change. The disk format is actually marked as stable and newer kernels will support older formats if a format change is necessary.

[darkbasic]

AFAIK even with the patches the write hole still exists, parity is still not checksummed and scrub will still not repair. It just fixes a scrub race condition which is not even mentioned in the wiki.

[timofonic]

Btrfs is smelling really rotten... Rancid butter?

[caligula]

AFAIK even with the patches the write hole still exists, parity is still not checksummed and scrub will still not repair. It just fixes a scrub race condition which is not even mentioned in the wiki.

On top of that it's also damn slow.

[ssam]

This is pretty good news. The worst case of "more or less fatally flawed, and a full scrap and rewrite to an entirely different raid56 mode on-disk format may be necessary to fix it" turns out not to be true. I'll probably stick with BTRFS raid 1 a bit longer though.

The write hole is not unique to BTRFS, and the checksumming reduces its impact. So I don't see that as a big deal.

[FireBurn]

If this is a bug fix, I'd expect it to go into 4.9 and be back ported to the stable kernels