[dm-devel] [PATCH] md/raid0: Fail BIOs if their underlying block device is gone

[Guilherme G. Piccoli]

On 29/07/2019 17:18, Roman Mamedov wrote:
> On Mon, 29 Jul 2019 16:33:59 -0300
> "Guilherme G. Piccoli" <gpiccoli at canonical.com> wrote:
> 
>> Currently md/raid0 is not provided with any mechanism to validate if
>> an array member got removed or failed. The driver keeps sending BIOs
>> regardless of the state of array members. This leads to the following
>> situation: if a raid0 array member is removed and the array is mounted,
>> some user writing to this array won't realize that errors are happening
>> unless they check kernel log or perform one fsync per written file.
>>
>> In other words, no -EIO is returned and writes (except direct ones) appear
>> normal. Meaning the user might think the wrote data is correctly stored in
>> the array, but instead garbage was written given that raid0 does stripping
>> (and so, it requires all its members to be working in order to not corrupt
>> data).
> 
> If that's correct, then this seems to be a critical weak point in cases when
> we have a RAID0 as a member device in RAID1/5/6/10 arrays.
> 

Hi Roman, I don't think this is usual setup. I understand that there are
RAID10 (also known as RAID 0+1) in which we can have like 4 devices, and
they pair in 2 sets of two disks using stripping, then these sets are
paired using mirroring. This is handled by raid10 driver however, so it
won't suffer for this issue.

I don't think it's common or even makes sense to back a raid1 with 2
pure raid0 devices.
Thanks for your comment!
Cheers,


Guilherme