how to set stride / raid-howto still up to date?

how to set stride / raid-howto still up to date?

[Dexter Filmore]

Want to change a partition from xfs to ext3 but can't tell what to put for 
stride.

man page says:

"stride=stripe-size
                          Configure  the  filesystem  for  a  RAID  array with
                          stripe-size filesystem blocks per stripe."

So, what is stripe size anyway? The same as chunk size? 

Is the example in the raid howto section 5.11 still valid? (It still tells 
about -R instead of -E. Seems a bit old.)

So going with -b 4096 for the ext3 with a 32k chunk size still comes down to a 
stride of 8, correct?

Dex

-- 
-----BEGIN GEEK CODE BLOCK-----
Version: 3.12
GCS d--(+)@ s-:+ a- C++++ UL++ P+>++ L+++>++++ E-- W++ N o? K-
w--(---) !O M+ V- PS+ PE Y++ PGP t++(---)@ 5 X+(++) R+(++) tv--(+)@ 
b++(+++) DI+++ D- G++ e* h>++ r* y?
------END GEEK CODE BLOCK------

http://www.stop1984.com
http://www.againsttcpa.com

Re: how to set stride / raid-howto still up to date?

[Mark Hahn]

> "stride=stripe-size
>                          Configure  the  filesystem  for  a  RAID  array with
>                          stripe-size filesystem blocks per stripe."

my understanding (hey, I even had a quick look at the source) is that
you want blocksize * stripe-size = raid-stripe-size, where the latter 
is chunksize * num-data-elements.

> So, what is stripe size anyway? The same as chunk size?

I don't know whether there's any agreed-on terminology.  mke2fs wants to know
how many fs-blocks it'll take to span a whole set of raid chunks.  that is, 
a write of X will hit all the disks in parallel.  writes that are full-stripe
(and aligned) don't require the read-mod-write cycle.

> So going with -b 4096 for the ext3 with a 32k chunk size still comes down to a
> stride of 8, correct?

no, I don't think so - I think it should be chunksize*ndatadisks/fsblocksize...

MegaRaid problems..

[Gordon Henderson]

This might not be strictly on-topic here, but you may provide
enlightenment, as a lot of web searching hasn't helpmed me so-far )-:

A client has bought some Dell hardware - Dell 1950 1U server, 2 on-board
SATA drives connected to a Fusion MPT SAS controller. This works just
fine. The on-board drives are mirrored using s/w RAID, which is great and
just how I want it.

The server also has 2 x Dell PERC dual-port SAS Raid Cards which have LSI
MegaRaid chipssets on them. One cable from each raid card connect to half
of a Dell external disk array box - 15 500GB SATA drives with a SAS
backplane, one card has 8 drives, the other 7. I want to run the RAID
cards in JBOD mode, so I can use linux s/w RAID. A nice little package,
which takes up 4U of rack space in total. (although the disk box is f'ing
heavy!!!)

And this is where I'm a little frustrated! I've compiled up a custom
kernel, (which is what I always do for my servers - no modules, no initrd,
this is 2.6.18), and at boot time the dmesg output sees the drives in the
external enclosure, but does not associate them to sdX drives! The
underlying distro is Debian stable, but I doubt theres anything of issue
there.

ie.:

megasas: 00.00.03.01 Sun May 14 22:49:52 PDT 2006
megasas: 0x1028:0x0015:0x1028:0x1f01: bus 12:slot 14:func 0
ACPI: PCI Interrupt 0000:0c:0e.0[A] -> GSI 18 (level, low) -> IRQ 66
scsi0 : LSI Logic SAS based MegaRAID driver
  Vendor: DELL      Model: MD1000            Rev: A.00
  Type:   Enclosure                          ANSI SCSI revision: 05
  Vendor: ATA       Model: HDS725050KLA360   Rev: AB5A
  Type:   Direct-Access                      ANSI SCSI revision: 05
  Vendor: ATA       Model: HDS725050KLA360   Rev: AB5A
  Type:   Direct-Access                      ANSI SCSI revision: 05
  Vendor: ATA       Model: HDS725050KLA360   Rev: AB5A
  Type:   Direct-Access                      ANSI SCSI revision: 05

etc. and it's repeated for the 2nd card. I've never seen this behaviour
before - normally it sees the drives, then associated them to sdX devices
- which is exactly what it does for the internal 2 drives:

On the internal drives, it goes like:

Fusion MPT base driver 3.04.01
Copyright (c) 1999-2005 LSI Logic Corporation
Fusion MPT SAS Host driver 3.04.01
ACPI: PCI Interrupt 0000:02:08.0[A] -> GSI 64 (level, low) -> IRQ 74
mptbase: Initiating ioc0 bringup
ioc0: SAS1068: Capabilities={Initiator}
scsi2 : ioc0: LSISAS1068, FwRev=00062800h, Ports=1, MaxQ=511, IRQ=74
  Vendor: ATA       Model: WDC WD1600JS-75N  Rev: 2E04
  Type:   Direct-Access                      ANSI SCSI revision: 05
SCSI device sda: 312500000 512-byte hdwr sectors (160000 MB)
sda: Write Protect is off
sda: Mode Sense: 67 00 00 08
SCSI device sda: drive cache: write back
SCSI device sda: 312500000 512-byte hdwr sectors (160000 MB)
sda: Write Protect is off
sda: Mode Sense: 67 00 00 08
SCSI device sda: drive cache: write back
 sda: sda1 sda2 sda3 sda4 < sda5 sda6 >
sd 2:0:0:0: Attached scsi disk sda
sd 2:0:0:0: Attached scsi generic sg2 type 0

and the same for the 2nd drive, sdb.

# cat/proc/scsi/scsi
Attached devices:
Host: scsi0 Channel: 00 Id: 00 Lun: 00
  Vendor: DELL     Model: MD1000           Rev: A.00
  Type:   Enclosure                        ANSI SCSI revision: 05
Host: scsi1 Channel: 00 Id: 00 Lun: 00
  Vendor: DELL     Model: MD1000           Rev: A.00
  Type:   Enclosure                        ANSI SCSI revision: 05
Host: scsi2 Channel: 00 Id: 00 Lun: 00
  Vendor: ATA      Model: WDC WD1600JS-75N Rev: 2E04
  Type:   Direct-Access                    ANSI SCSI revision: 05
Host: scsi2 Channel: 00 Id: 01 Lun: 00
  Vendor: ATA      Model: WDC WD1600JS-75N Rev: 2E04
  Type:   Direct-Access                    ANSI SCSI revision: 05

So it's a bit frustrating.

LUN support is enabled, and the boot-line is:

  auto BOOT_IMAGE=Linux ro root=901 max_luns=8

So if it's a LUN issue, then they ought to be being probed...

I'm wondering about things like the controllers needing some poking at the
BIOS level, but I've looked and there isn't a JBOD mode - only various
built-in RAID modes, so I've not created any RAID sets through the BIOS (I
want raid 6 on this box over all 15 drives)

I'm sure it's something dead obvious, so any clues would be appreciated!

Thanks,

Gordon

Re: MegaRaid problems..

[Joshua Baker-LePain]

On Thu, 12 Oct 2006 at 10:06am, Gordon Henderson wrote

> I'm wondering about things like the controllers needing some poking at the
> BIOS level, but I've looked and there isn't a JBOD mode - only various
> built-in RAID modes, so I've not created any RAID sets through the BIOS (I
> want raid 6 on this box over all 15 drives)

Some RAID cards don't explicitly have a JBOD mode but instead an option 
that says something like "Export unconfigured disks", and ISTR that 
MegaRAID may be that sort of card.  Poke about in the card BIOS a bit 
more.

-- 
Joshua Baker-LePain
Department of Biomedical Engineering
Duke University

Re: MegaRaid problems..

[Mike Hardy]

Gordon Henderson wrote:
> This might not be strictly on-topic here, but you may provide
> enlightenment, as a lot of web searching hasn't helpmed me so-far )-:
> 
> A client has bought some Dell hardware - Dell 1950 1U server, 2 on-board
> SATA drives connected to a Fusion MPT SAS controller. This works just
> fine. The on-board drives are mirrored using s/w RAID, which is great and
> just how I want it.
> 
> The server also has 2 x Dell PERC dual-port SAS Raid Cards which have LSI
> MegaRaid chipssets on them. One cable from each raid card connect to half


> this is 2.6.18), and at boot time the dmesg output sees the drives in the
> external enclosure, but does not associate them to sdX drives! The
> underlying distro is Debian stable, but I doubt theres anything of issue
> there.

I have several Dell 2950s (same chassis) and they have this problem.

You can't do the PERC card and get JBOD basically. The PERC5 card has no
JBOD mode, whereas the PERC4 card did.

Dell said they may get a BIOS update, but wouldn't commit.

In the meantime, you have to exchange the PERC5 card for a SAS5 card,
then you can have JBOD.

I was a little disappointed, as the PERC5 card can drive 6 or 8 devices,
but the SAS5 card can only drive 4. Lame.

-Mike