non-fakeraid controllers

non-fakeraid controllers

[alex-lists-linux-kernel]

Hi,

	This is not an attempt to start a religious flamewar about what is
RAID vs. what is softraid vs. what is fakeraid. 

	Does anyone has a list/refence/etc on reasonably modern SCSI
controllers (at least u160) in a non-fakeraid way i.e. the way that would
allow linux to boot from a RAID protected disk array when one of the drives
in the array failed even if the root filesystem is located on the same
array?

Thanks,
Alex

Re: non-fakeraid controllers

[Helge Hafting]

alex-lists-linux-kernel@yuriev.com wrote:

>Hi,
>
>	This is not an attempt to start a religious flamewar about what is
>RAID vs. what is softraid vs. what is fakeraid. 
>
>	Does anyone has a list/refence/etc on reasonably modern SCSI
>controllers (at least u160) in a non-fakeraid way i.e. the way that would
>allow linux to boot from a RAID protected disk array when one of the drives
>in the array failed even if the root filesystem is located on the same
>array?
>  
>
Ability to boot requires bios/bootloader support. Depending on the bios in
question, this may work for real RAID, fakeraid or even linux sw raid.

You can boot directly from the software raid-1 in linux.  And you can set
it up so it will boot with one drive failed too - although you may have
to disconnect the bad drive so the biosdoesn't mistakenly try to load the
kernel bootloader from the damaged disk.

Having the root filesytem on a damaged array is not a problem - as soon
as the kernel is running it can activate the raid in degraded mode and
use the filesystem just fine.  This is no more secure than a single-disk 
setup
though, so don't wait too long before you replace the failed drive.

Fakeraid controllers may have bios support for booting, but often
you'll find that linux have no support for the fake raid.  So you have
to turn that off and use software raid instead.  An expensive "real raid"
controller that have linux support, will usually have a bios that support
booting from the raid too.  Writing to manufacturers should get you the
details on booting in degraded conditions.

Helge Hafting

Re: non-fakeraid controllers

[Phillip Susi]

Helge Hafting wrote:
> Ability to boot requires bios/bootloader support. Depending on the 
> bios in
> question, this may work for real RAID, fakeraid or even linux sw raid.
>
> You can boot directly from the software raid-1 in linux.  And you can set
> it up so it will boot with one drive failed too - although you may have
> to disconnect the bad drive so the biosdoesn't mistakenly try to load the
> kernel bootloader from the damaged disk.
>
> Having the root filesytem on a damaged array is not a problem - as soon
> as the kernel is running it can activate the raid in degraded mode and
> use the filesystem just fine.  This is no more secure than a 
> single-disk setup
> though, so don't wait too long before you replace the failed drive.
>
> Fakeraid controllers may have bios support for booting, but often
> you'll find that linux have no support for the fake raid.  So you have
> to turn that off and use software raid instead.  An expensive "real raid"
> controller that have linux support, will usually have a bios that support
> booting from the raid too.  Writing to manufacturers should get you the
> details on booting in degraded conditions.
>


Heinz wrote a utility called 'dmraid' which detects and activates 
various hardware fakeraid volumes using the kernel device mapper.  I'm 
using it at home to dual boot ubuntu and winxp ( just in case ) from a 
stripe of two WD 10,000 rpm raptors and it works great.  I'm still 
pushing to get it integrated into dapper.  At this time however, neither 
my hardware, nor dmraid support raid-5.  Hopefully this will change in 
the future. 

Re: non-fakeraid controllers

[Wakko Warner]

Phillip Susi wrote:
> Heinz wrote a utility called 'dmraid' which detects and activates 
> various hardware fakeraid volumes using the kernel device mapper.  I'm 
> using it at home to dual boot ubuntu and winxp ( just in case ) from a 
> stripe of two WD 10,000 rpm raptors and it works great.  I'm still 
> pushing to get it integrated into dapper.  At this time however, neither 
> my hardware, nor dmraid support raid-5.  Hopefully this will change in 
> the future. 

I've been wondering about dmraid.  I considered buying an adaptec sata raid
(hardware).  One of the drawbacks on hardware raid is the format isn't
compatible with any other card (or rather manufacturer).  So the question
is, has anyone written anything that can detect and activate disks from a
hardware raid controller when they are placed on a controller w/o any raid?

basically what I mean is, 3 disks, raid5 was in a system with hardware raid. 
the raid card blows up and cannot get another one so to get the data back,
place disks in another machine or on a standard controller and use software
raid or whatever to recover the data.

-- 
 Lab tests show that use of micro$oft causes cancer in lab animals
 Got Gas???

Re: non-fakeraid controllers

[Phillip Susi]

Wakko Warner wrote:
> I've been wondering about dmraid.  I considered buying an adaptec sata raid
> (hardware).  One of the drawbacks on hardware raid is the format isn't
> compatible with any other card (or rather manufacturer).  So the question
> is, has anyone written anything that can detect and activate disks from a
> hardware raid controller when they are placed on a controller w/o any raid?
>
> basically what I mean is, 3 disks, raid5 was in a system with hardware raid. 
> the raid card blows up and cannot get another one so to get the data back,
> place disks in another machine or on a standard controller and use software
> raid or whatever to recover the data.

Yes, that is one of the problems with hardware raid controllers.  If you 
can figure out the controller's metadata format, possibly by asking 
adaptec, or reverse engineering, then patching dmraid to understand that 
format ( it already understands several used by fakeraid controllers ) 
should be rather easy. 

Re: non-fakeraid controllers

[Wakko Warner]

Phillip Susi wrote:
> Wakko Warner wrote:
> >I've been wondering about dmraid.  I considered buying an adaptec sata raid
> >(hardware).  One of the drawbacks on hardware raid is the format isn't
> >compatible with any other card (or rather manufacturer).  So the question
> >is, has anyone written anything that can detect and activate disks from a
> >hardware raid controller when they are placed on a controller w/o any raid?
> >
> >basically what I mean is, 3 disks, raid5 was in a system with hardware 
> >raid. the raid card blows up and cannot get another one so to get the data 
> >back,
> >place disks in another machine or on a standard controller and use software
> >raid or whatever to recover the data.
> 
> Yes, that is one of the problems with hardware raid controllers.  If you 
> can figure out the controller's metadata format, possibly by asking 
> adaptec, or reverse engineering, then patching dmraid to understand that 
> format ( it already understands several used by fakeraid controllers ) 
> should be rather easy. 

One of the things that I noticed with adaptec's and ami's raid (and an old mylex
dac960 IIRC) is that you can create containers or partitions (which every
they call them) on a disk set.  I guess you can take 3 disks, create a raid5
using half the capcity of the 3 disks and a raid0 on the rest.  I'm not sure
about fakeraid controllers as I've never configured one.

-- 
 Lab tests show that use of micro$oft causes cancer in lab animals
 Got Gas???

Re: non-fakeraid controllers

[Christoph Hellwig]

On Mon, Feb 06, 2006 at 08:51:26PM -0500, alex-lists-linux-kernel@yuriev.com wrote:
> Hi,
> 
> 	This is not an attempt to start a religious flamewar about what is
> RAID vs. what is softraid vs. what is fakeraid. 
> 
> 	Does anyone has a list/refence/etc on reasonably modern SCSI
> controllers (at least u160) in a non-fakeraid way i.e. the way that would
> allow linux to boot from a RAID protected disk array when one of the drives
> in the array failed even if the root filesystem is located on the same
> array?

LSI 1030/1035 U320 (fusion) controllers have simple raid0/raid1 support

Adaptec and LSI still have some u160 or even u320 controllers in the aacraid/
megaraid series afaik, the present the same interface to the OS for parallel
scsi/sata/sas so it's a bit hard to say for me which is the most recent
parallel scsi one.

The old Mylex controllers supported by drivers/block/DAC960.c support up
to u160, and u320 with an IBM-branded controller which probably isn't sold
separately from IBM Equipment and probably not at all anymore.

The IBM i/pSeries integrated RAID supports up to U320 (drivers/scsi/ipr.c),
but you don't get it without an i/pSeries system.

There's Intel Branded, Adaptec manufactured RAID cards that support U160
SCSI, they're supported by drivers/scsi/gdth.c

Re: non-fakeraid controllers

[Alan Cox]

On Llu, 2006-02-06 at 20:51 -0500, alex-lists-linux-kernel@yuriev.com
wrote:
> 	Does anyone has a list/refence/etc on reasonably modern SCSI
> controllers (at least u160) in a non-fakeraid way i.e. the way that would
> allow linux to boot from a RAID protected disk array when one of the drives
> in the array failed even if the root filesystem is located on the same
> array?

Most raid (soft or otherwise) will usually do it. I don't know any SCSI
hardware which will do it reliably unless you go fibrechannel or SATA
simply because a dead scsi drive can and sometimes does hang the entire
shared bus.

The majority of the time the aacraid based cards will do what you need
(most of the time anyway) and can also do a lot of on the fly recovery
and management. The fusions can do some raid stuff of this nature. Most
of the newer hardware is going SATA however.

Alan

Re: non-fakeraid controllers

[Jesper Juhl]

On 2/7/06, alex-lists-linux-kernel@yuriev.com
<alex-lists-linux-kernel@yuriev.com> wrote:
> Hi,
>
>         This is not an attempt to start a religious flamewar about what is
> RAID vs. what is softraid vs. what is fakeraid.
>
>         Does anyone has a list/refence/etc on reasonably modern SCSI
> controllers (at least u160) in a non-fakeraid way i.e. the way that would
> allow linux to boot from a RAID protected disk array when one of the drives
> in the array failed even if the root filesystem is located on the same
> array?
>

Personally I have very good exeriences with Linux & IBM ServeRAID
controllers (especially the ServeRAID-5i, ServeRAID-5M & ServeRAID-6M)
as well as HP's Integrated Smart Array 6i controller in their ProLiant
DL380 G4 servers.

--
Jesper Juhl <jesper.juhl@gmail.com>
Don't top-post  http://www.catb.org/~esr/jargon/html/T/top-post.html
Plain text mails only, please      http://www.expita.com/nomime.html