[dm-devel] Proper way to test RAID456?

[Qu Wenruo]

On 2022/1/9 04:29, Lukas Straub wrote:
> On Sat, 8 Jan 2022 19:52:59 +0000
> Lukas Straub <lukasstraub2 at web.de> wrote:
>
>> CC'ing linux-raid mailing list, where md raid development happens.
>> dm-raid is just a different interface to md raid.
>>
>> On Fri, 7 Jan 2022 10:30:56 +0800
>> Qu Wenruo <quwenruo.btrfs at gmx.com> wrote:
>>
>>> Hi,
>>>
>>> Recently I'm working on refactor btrfs raid56 (with long term objective
>>> to add proper journal to solve write-hole), and the coverage of current
>>> fstests for btrfs RAID56 is not that ideal.
>>>
>>> Is there any project testing dm/md RAID456 for things like
>>> re-silvering/write-hole problems?
>>>
>>> And how you dm guys do the tests for stacked RAID456?
>>>
>>> I really hope to learn some tricks from the existing, tried-and-true
>>> RAID456 implementations, and hopefully to solve the known write-hole
>>> bugs in btrfs.
>
> Just some thoughts:
> Besides the journal to mitigate the write-hole, md raid has another
> trick:
> The Partial Parity Log
> https://www.kernel.org/doc/html/latest/driver-api/md/raid5-ppl.html
>
> When a stripe is partially updated with new data, PPL ensures that the
> old data in the stripe will not be corrupted by the write-hole. The new
> data on the other hand is still affected by the write hole, but for
> btrfs that is no problem.
>
> But there is a even simpler solution for btrfs: It could just not touch
> stripes that already contain data.

That would waste a lot of space, if the fs is fragemented.

Or we have to write into data stripes when free space is low.

That's why I'm trying to implement a PPL-like journal for btrfs RAID56.

Thanks,
Qu

>
> The big problem will be NOCOW files, since a write to an already
> allocated extent will necessarily touch a stripe with old data in it
> and the new data also needs to be protected from the write-hole.
>
> Regards,
> Lukas Straub
>
>>> Thanks,
>>> Qu
>>>
>>>
>>> --
>>> dm-devel mailing list
>>> dm-devel at redhat.com
>>> https://listman.redhat.com/mailman/listinfo/dm-devel
>>>
>>
>>
>>
>
>
>