[linux-lvm] raid and lvm

SI Reasoning si at si.homelinux.org
Tue Jan 28 12:55:02 UTC 2003

maybe I am misunderstanding. I was looking to do software raid (off of a non-raid
SCSI controller.)
The main reason I wanted raid5 was as a failsafe measure since it would appear
that things can get tricky if a harddrive fails in lvm. Is there another way to
create a failsafe through lvm without needing to implement software raid?

Christophe Saout (christophe at saout.de) wrote*:
>Am Die, 2003-01-28 um 18.34 schrieb SI Reasoning:
>> I was following an old list and at that time they felt that it would be
>> more efficient to merge either lvm into raid or vice-versa.
>Yes. I've looked into the DM code in the kernel and I have to say it's a
>very clean architecture. Since LVM2 needs some kind of Raid 0 (a
>transparent mirror in the background) to implement pvmove it would be a
>code duplication.
>Also since DM has a nice plugin architecture and already has generic bio
>splitting code, it would even save some memory when the raid drivers
>would be reimplemented as DM plugins. (so the bloat argument doesn't
>Also you don't need the extra meta data the raid code actually has (ok,
>raid autodetection won't work anymore but as far I understand the kernel
>developers, that's depracted anyway). I've had problems with the raid
>metadata code that could have been solved with things like vgexport and
>But I think that's going to be adressed via the new in-kernel
>initramdisk anyway.
>And I think that the performance is better when raid is implemented in
>DM than to run one device mapper over another one.
>linux-lvm mailing list
>linux-lvm at sistina.com
>read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/

SI Reasoning
si at mindspring.com
gpg public key ftp://ftp.p-p-i.com/pub/si-mindspring-pubkey.asc

The significant problems we face cannot be solved by
the same level of thinking that created them.
-Albert Einstein

More information about the linux-lvm mailing list