[linux-lvm] Best way to run LVM over multiple SW RAIDs?

[Guoqing Jiang]

On 12/7/19 11:44 PM, John Stoffel wrote:
>>>>>> "Stuart" == Stuart D Gathman<stuart at gathman.org>  writes:
> Stuart> On Tue, Oct 29, 2019 at 12:14 PM Daniel Janzon<daniel.janzon at edgeware.tv>  wrote:
>>> I have a server with very high load using four NVMe SSDs and
>>> therefore no HW RAID. Instead I used SW RAID with the mdadm tool.
>>> Using one RAID5 volume does not work well since the driver can only
>>> utilize one CPU core which spikes at 100% and harms performance.
>>> Therefore I created 8 partitions on each disk, and 8 RAID5s across
>>> the four disks.
>>> Now I want to bring them together with LVM. If I do not use a striped
>>> volume I get high performance (in expected magnitude according to disk
>>> specs). But when I use a striped volume, performance drops to a
>>> magnitude below. The reason I am looking for a striped setup is to
> Stuart> The mdadm layer already does the striping.  So doing it again
> Stuart> in the LVM layer completely screws it up.  You want plain JBOD
> Stuart> (Just a Bunch Of Disks).
>
> Umm... not really.  The problem here is more the MD layer not being
> able to run RAID5 across multiple cores at the same time, which is why
> he split things the way he did.
>
> But we don't know the Kernel version, the LVM version, or the OS
> release so as to give better ideas of what to do.
>
> The biggest harm to performance here is really the RAID5, and if you
> can instead move to RAID 10 (mirror then stripe across mirrors) then
> you should be a performance boost.
>
> As Daniel says, he's got lots of disk load, but plenty of CPU, so the
> single thread for RAID5 is a big bottleneck.

Perhaps set "/sys/block/mdx/md/group_thread_cnt" could help here,
see below commits:

commit b721420e8719131896b009b11edbbd27d9b85e98
Author: Shaohua Li <shli at kernel.org>
Date:   Tue Aug 27 17:50:42 2013 +0800

     raid5: sysfs entry to control worker thread number

commit 851c30c9badfc6b294c98e887624bff53644ad21
Author: Shaohua Li <shli at kernel.org>
Date:   Wed Aug 28 14:30:16 2013 +0800

     raid5: offload stripe handle to workqueue

Thanks,
Guoqing