Btrfs Will Finally "Strongly Discourage" You When Creating RAID5 / RAID6 Arrays

jugganuts

https://phoronix.com/scan.php?page=news_item&px=Btrfs-Warning-RAID5-RAID6

I use ZFS because of issues like these. no regrets.

TimboJones

Then you have not been reading the stories of losing ZFS arrays.

darkimmortal

The ‘bug’ (more incomplete implementation) with raid 5/6 is well known at this point, and raid 5/6 are a dying breed with modern large disks. Doesn’t seem like a particularly good reason to choose ZFS.

It’s an interesting situation where the public perception might be better if they didn’t support raid 5/6 at all. You see the same thing in video games, developers intentionally nerf capabilities so that people can feel better about themselves when their systems can run games on Ultra.

If you want a good reason to think twice about BTRFS, the ‘bug’ with BTRFS raid in conjunction with nodatacow/chattr +C is at least as bad and less well known.

At present, nodatacow implies nodatasum; this means that anything with the nodatacow attribute does not receive the benefits of btrfs' checksum protection and self-healing (for raid levels greater >= 1); disabling CoW (Copy on Write) means that the a VM disk image will not be consistent if the host crashes or loses power. Nodatacow also carries the following additional danger on multidisk systems: because nodatasum is disabled there is no way to verify which disk in a two disk raid1 profile volume contains the correct data. After a crash there is a roughly 50% probability that the bad copy will be read on each request. Consequently, it is almost always preferable to disable COW in the application, and nodatacow should only be used for disposable data

Source: https://wiki.debian.org/Btrfs

And I say this as a happy btrfs user - great filesystem when you know its limits, and I prefer its design and tools to ZFS. I’m a bit surprised they chose to advertise the well known raid 5/6 issue and not the less well known raid+nodatacow issue

notarobo

there should be no other raid then raid 10.

jugganuts

@TimboJones said:
Then you have not been reading the stories of losing ZFS arrays.

I have not most what iv heard is ZFS updating but DKMS module is lagging behind and causing issues.
or where people use the FUSE module

My oldest array is 5ish years strong now. no hickups besides updating.

Daniel15

Do many people even use software RAID5 or 6? I thought people that use those RAID levels primarily use hardware RAID...

@darkimmortal said: the ‘bug’ with BTRFS raid in conjunction with nodatacow/chattr +C is at least as bad and less well known.

Why would someone use a copy-on-write file system and then explicitly disable copy-on-write? 🤔

Xsltel

@Daniel15 said: hardware RAID

they fail and fail so hard to the point you won't find a controller to replace them. hardware RAID left a very bad taste in my mouth a few months ago.

MINIMAN10000

@notarobo said:
there should be no other raid then raid 10.

Isn't the whole idea behind striped arrays is the ability to lose any n drives where n is the number of drives being used as striped storage.

Raid 10 if you lose the wrong pair of drives the data becomes lost.

Wouldn't it be better to be able to lose 2 any drives?