two physical drives, but two device files point to same drive

two physical drives, but two device files point to same drive

[Matt Garman]

I have a backup server with 25 total SATA drives.  I just swapped two
drives for two new ones.  When I boot, there are two /dev/sdX device
files, but both device files actually refer to the same physical
drive.  The other physical drive does not have a device file and
cannot be accessed.

Here is a dmesg excerpt where the two drives are identified:

mptsas: ioc0: attaching sata device: fw_channel 0, fw_id 21, phy 26, sas_addr 0x
50030480006cd85a
  Vendor: ATA       Model: Hitachi HDS72202  Rev: A3EA
  Type:   Direct-Access                      ANSI SCSI revision: 05
SCSI device sdx: 3907029168 512-byte hdwr sectors (2000399 MB)
sdx: Write Protect is off
sdx: Mode Sense: 73 00 00 08
SCSI device sdx: drive cache: write back
SCSI device sdx: 3907029168 512-byte hdwr sectors (2000399 MB)
sdx: Write Protect is off
sdx: Mode Sense: 73 00 00 08
SCSI device sdx: drive cache: write back
 sdx: sdx1
sd 4:0:22:0: Attached scsi disk sdx
mptsas: ioc0: attaching sata device: fw_channel 0, fw_id 21, phy 27,
sas_addr 0x50030480006cd85b
  Vendor: ATA       Model: Hitachi HDS72202  Rev: A3EA
  Type:   Direct-Access                      ANSI SCSI revision: 05
SCSI device sdy: 3907029168 512-byte hdwr sectors (2000399 MB)
sdy: Write Protect is off
sdy: Mode Sense: 73 00 00 08
SCSI device sdy: drive cache: write back
SCSI device sdy: 3907029168 512-byte hdwr sectors (2000399 MB)
sdy: Write Protect is off
sdy: Mode Sense: 73 00 00 08
SCSI device sdy: drive cache: write back
 sdy: sdy1
sd 4:0:23:0: Attached scsi disk sdy

So the two drives are supposed to be sdx and sdy.

Here is an excerpt from "smartctl -a /dev/sdx":

Device Model:     Hitachi HDS722020ALA330
Serial Number:    JK11A1YAJHEX7V
Firmware Version: JKAOA3EA
User Capacity:    2,000,398,934,016 bytes


Likewise, "smartctl -a /dev/sdy":

Device Model:     Hitachi HDS722020ALA330
Serial Number:    JK11A1YAJHEX7V
Firmware Version: JKAOA3EA
User Capacity:    2,000,398,934,016 bytes

As you can see, the serial numbers are the same.

Furthermore, the drives are both brand new, and come from the factory
without any partition table.  I created a partition table and ext3
filesystem on sdx.  When I started fdisk for sdy, a partition already
existed.  I went ahead and tried to create an ext3 filesystem on sdy
as well, but after creating the inode blocks, it failed, saying, "file
already exists".

The drives are attached to an LSI Logic SASX36 expander/backplane.
The SAS controller is: LSI Logic / Symbios Logic SAS1068E PCI-Express
Fusion-MPT SAS.

The kernel is from RHEL/CentOS: 2.6.18-194.3.1.el5

With the exception of these two new Hitachi drives, I have been using
this hardware for several months, with other drives connected to the
exact same expander ports.  The only difference is that this is the
first time I've used Hitachi drives.  Is it possible they are not
compatible with the expander?  Or is it a kernel/driver bug (for
either the SAS controller or expander)?

Thanks,
Matt

Re: two physical drives, but two device files point to same drive

[Matt Garman]

Sorry to post this before I did more investigating.  Turns out one of
the new Hitachi drives was bad.

Still, think it's a curious failure model though.  If I pulled the
drive that was "doubly" recognized, then the bad drive isn't
recognized by Linux at all.  If I pulled the bad drive, the good drive
is no longer doubly recognized (i.e. has exactly one device file).

Strange.  But, replacing the broken drive with a good one makes things
look and act like they should.

Thanks,
Matt

On Wed, Sep 1, 2010 at 10:18 AM, Matt Garman <matthew.garman@gmail.com> wrote:
> I have a backup server with 25 total SATA drives.  I just swapped two
> drives for two new ones.  When I boot, there are two /dev/sdX device
> files, but both device files actually refer to the same physical
> drive.  The other physical drive does not have a device file and
> cannot be accessed.
>
> Here is a dmesg excerpt where the two drives are identified:
>
> mptsas: ioc0: attaching sata device: fw_channel 0, fw_id 21, phy 26, sas_addr 0x
> 50030480006cd85a
>  Vendor: ATA       Model: Hitachi HDS72202  Rev: A3EA
>  Type:   Direct-Access                      ANSI SCSI revision: 05
> SCSI device sdx: 3907029168 512-byte hdwr sectors (2000399 MB)
> sdx: Write Protect is off
> sdx: Mode Sense: 73 00 00 08
> SCSI device sdx: drive cache: write back
> SCSI device sdx: 3907029168 512-byte hdwr sectors (2000399 MB)
> sdx: Write Protect is off
> sdx: Mode Sense: 73 00 00 08
> SCSI device sdx: drive cache: write back
>  sdx: sdx1
> sd 4:0:22:0: Attached scsi disk sdx
> mptsas: ioc0: attaching sata device: fw_channel 0, fw_id 21, phy 27,
> sas_addr 0x50030480006cd85b
>  Vendor: ATA       Model: Hitachi HDS72202  Rev: A3EA
>  Type:   Direct-Access                      ANSI SCSI revision: 05
> SCSI device sdy: 3907029168 512-byte hdwr sectors (2000399 MB)
> sdy: Write Protect is off
> sdy: Mode Sense: 73 00 00 08
> SCSI device sdy: drive cache: write back
> SCSI device sdy: 3907029168 512-byte hdwr sectors (2000399 MB)
> sdy: Write Protect is off
> sdy: Mode Sense: 73 00 00 08
> SCSI device sdy: drive cache: write back
>  sdy: sdy1
> sd 4:0:23:0: Attached scsi disk sdy
>
> So the two drives are supposed to be sdx and sdy.
>
> Here is an excerpt from "smartctl -a /dev/sdx":
>
> Device Model:     Hitachi HDS722020ALA330
> Serial Number:    JK11A1YAJHEX7V
> Firmware Version: JKAOA3EA
> User Capacity:    2,000,398,934,016 bytes
>
>
> Likewise, "smartctl -a /dev/sdy":
>
> Device Model:     Hitachi HDS722020ALA330
> Serial Number:    JK11A1YAJHEX7V
> Firmware Version: JKAOA3EA
> User Capacity:    2,000,398,934,016 bytes
>
> As you can see, the serial numbers are the same.
>
> Furthermore, the drives are both brand new, and come from the factory
> without any partition table.  I created a partition table and ext3
> filesystem on sdx.  When I started fdisk for sdy, a partition already
> existed.  I went ahead and tried to create an ext3 filesystem on sdy
> as well, but after creating the inode blocks, it failed, saying, "file
> already exists".
>
> The drives are attached to an LSI Logic SASX36 expander/backplane.
> The SAS controller is: LSI Logic / Symbios Logic SAS1068E PCI-Express
> Fusion-MPT SAS.
>
> The kernel is from RHEL/CentOS: 2.6.18-194.3.1.el5
>
> With the exception of these two new Hitachi drives, I have been using
> this hardware for several months, with other drives connected to the
> exact same expander ports.  The only difference is that this is the
> first time I've used Hitachi drives.  Is it possible they are not
> compatible with the expander?  Or is it a kernel/driver bug (for
> either the SAS controller or expander)?
>
> Thanks,
> Matt
>