Btrfs RAID 5/6 Code Found To Be Very Unsafe & Will Likely Require A Rewrite

[phoronix]

Phoronix: Btrfs RAID 5/6 Code Found To Be Very Unsafe & Will Likely Require A Rewrite

It turns out the RAID5 and RAID6 code for the Btrfs file-system's built-in RAID support is faulty and users should not be making use of it if you care about your data...

http://www.phoronix.com/scan.php?pag...RAID-56-Is-Bad

[johnc]

My friend recently bought a Synology NAS and I was surprised that they were recommending he select btrfs during its setup. And he wanted to do RAID5, too.

[michal]

"It turns out the RAID5 and RAID6 code for the Btrfs file-system's built-in RAID support is faulty and users should not be making use of it if you care about your data. "

that's funny, because usually you want to use raid if you care about your data. anyway, btrfs is still marked as experimental, so you should not use it if you care about your data. I wonder when developers will declare it stable.

[PuckPoltergeist]

"It turns out the RAID5 and RAID6 code for the Btrfs file-system's built-in RAID support is faulty and users should not be making use of it if you care about your data. "

that's funny, because usually you want to use raid if you care about your data. anyway, btrfs is still marked as experimental, so you should not use it if you care about your data. I wonder when developers will declare it stable.

The experimental mark was removed 2013.

RAID5/6 is still not stable and beside this new warning, it was recommended to use it only for testing all the time.

[Hohlraum]

Apparently I'm a pleb. I still use plain old linux software raid when the machine doesn't have full hardware raid.

[gbcox]

OK, what exactly is going on here? I was under the impression that RAID5/6 had been given a green light for regular usage - and now everyone seems to be in a panic. I've been using raid6 for 3 years now with no issues. In any event, for ALL kinds of RAID the recommendation is that it should NEVER be used as a backup. That said, if there is a problem the BTRFS folks need to jump on this and fix it immediately and NOT let it flounder. I've been a big proponent of BTRFS but enough is enough. If this is indeed suppose to be the next linux filesystem it needs to be taken seriously and not as some type of hobby project. RAID5/6 is a big part of that. This apparent indecisiveness and lack of resources is just ridiculous for a project that is "suppose" to be strategic. How many years has this thing been in development?

[Brane215]

So much for the "FS of the future".

For me, most ironic by far is their slogan "don't do your RAID, you must leave it to us" only to screw it royally and usually unrecoverably. RAID integration wasn't some peripheral bulletpoint, it was to be central feature.

Maybe they should change the name to ByTeschRedder FS or something...

I thought Reiser's jailtime would be game over for him, but the way BTRFS crowd is advancing, he could easily do one or two killing escapades and still come ahead with Reiser{4..9}.

[ssam]

I don't think anyone has ever claimed that raid5/6 in btrfs was ready for general use.

Its been known for a while that there are still issues with recovering from a failed disk in raid5/6 mode, and this seems to be the first hint at where the problem might be. So in a way its good news that they finally have a handle on where the bug is, which is the first step in fixing it. Finding the bug is the most difficult part of debugging, especially in a case like this where there is a big delay between the bad parity being written and you noticing it.

[Tomin]

For some reason you are missing an e and therefore the tag doesn't work as it's supposed to:

The <a href="https://btrfs.wiki.kernel.org/index.php/RAID56">RAID56 wiki page</a> has already been updated so users don't accidentally try one of these Btrfs RAID levels.

[ssam]

OK, what exactly is going on here? I was under the impression that RAID5/6 had been given a green light for regular usage

Where did you get that impression? btrfs raid5/6 only gained the ability to scrub and replace a failed a drive in 2015. A major new feature in a filesystem should be given at least a year (probably more like 3) before you'd consider it to have had a decent amount of testing. And from the mailing list it has been fairly clear that there are still some serious bugs lurking in that code. No one should be surprised at that.