why the disk shows as /dev/cciss/c0d0pX instead of /dev/sdaX ?

[Maaz]

Machine: HP Proliant ML370 G5

# df -h
Filesystem Size Used Avail Use% Mounted on
/dev/cciss/c0d0p2 502M 39M 464M 8% /boot
/dev/cciss/c0d0p3 443G 388G 55G 88% /

SCSI hard drives are /dev/sdX, but as per the output its shows /dev/cciss/c0d0pX why ?

Regards
Maaz

[Avijit Patra · Accepted solution]

The server has hardware raid and the cciss driver is showing the logicaldrive as /dev/cciss/c0d0pX. You have 2 slices on the c0d0.

[Ivan Ferreira]

That is normal and device names can change depending on the hardware module/driver. It's the driver provider desition :)

[Maaz]

Thanks Avijit Patra
>The server has hardware raid .....
If the we doesnt configure the hw raid, then it will show us like /dev/sdX instead of /dev/cciss/c0d0pX .. right ?
in other words /dev/cciss module/driver shows that hw raid is configured... am i right ?


and does c0d0 means Controller 0, Disk 0 ?

[Yogeeraj_1]

hi Maaz,

> If the we doesnt configure the hw raid, then it will show us like /dev/sdX instead of /dev/cciss/c0d0pX .. right ?
in other words /dev/cciss module/driver shows that hw raid is configured... am i right ?

You have a unique configuration and the behaviour or device naming may be different on another system running the same Operating system. Basically, it follows a specific naming convention.


>and does c0d0 means Controller 0, Disk 0 ?
Yes. You are right. Note also that there exist systems in which two controllers can point to the same disk, so you will have 2 device names for the device...

More recently, multipath driver which is intelligent enough to do the abstraction such that one device file points to one "physical" disk through 2 controllers.

Hope this is not too confusing.


All the best!

kind regards
yogeeraj

[Heironimus]

The /dev/cciss/ device names only mean that you have a SmartArray controller. Anything attached to a SmartArray controller will show up under that path, even if you just configure single drives.

[Asmer]

So there is no way to change this to /dev/sdx?

[Ivan Ferreira]

Any reason do do that? You could create alternate names with UDEV.

[Ross Minkov]

Maaz,

I wouldn't change change anything. Just use the device names the way they are.

Again, you are using the cciss driver as you have a hp Smart Array controller in your hp ProLiant server. All Smart Array controllers use the cciss driver. Red Hat has one that comes with their distro and then HP has an improved version that comes with the ProLiant Support Pack. I'd recommend that you install the ProLiant Support Pack on your ProLiant server.

-Ross

[Ross Minkov]

Some more info on the HP cciss driver -- it's called cpq_cciss -- HP ProLiant Smart Array Controller Driver.

This RPM is used to provide binary driver modules that have been pre-built to enable ProLiant Smart Array SAS/SATA, 6x and 5x Series controllers. This RPM provides a customer friendly approach to install a pre-built binary driver. On some kernel erratas the version of the driver present in the kernel may be equal to the version being installed by the rpm. However, the driver in the kernel may not include all of the fixes and enhancements implemented in the driver contained within the rpm. Therefore the rpm should be installed in order to obtain the fixes and enhancements that are not present in the kernel driver.


# rpm -qi cpq_cciss
Name : cpq_cciss Relocations: /opt/hp/storage_drivers
Version : 2.4.60 Vendor: Hewlett-Packard Company
Release : 13 Build Date: Wed 11 Jan 2006 10:36:58 AM CST
Install Date: Tue 20 Jun 2006 10:17:16 AM CDT Build Host: ld-rhel3a
Group : System Environment/Kernel Source RPM: cpq_cciss-2.4.60-13.src.rpm
Size : 547287 License: GPL
Signature : (none)
Packager : Hewlett-Packard Company
URL : http://www.hp.com/linux
Summary : HP cciss 2.4.60 HBA driver RPM
Description :
This RPM will install the Hewlett-Packard Company cpq_cciss
version 2.4.60-13 driver.

It will install a gzipped tar file, cciss-2.4.60_2006_01_11.tar.gz
into /opt/hp/storage_drivers. (The destination directory may be changed
vi RPM's --prefix or --relocate options, provided you have rpm
v. 4.0.3 or better. Additionally, by default, it will install the
driver object modules under the appropriate /lib/modules subdirectory,
attempt to create a new initrd image, and modify your LILO or GRUB
configuration. If you do no wish to have your LILO or GRUB
configuration or initrd image modified, you have several options.

You can install the RPM with the "--noscripts" option, which will
suppress all actions except for installation of the gzipped tar
file, so, LILO and GRUB configurations will not be disturbed and a
new initrd image will not be created. Then, you can unpack the
gzipped tar file and install the drivers yourself using the "install"
script contained within the tar archive. (Documentation for the
"install" script is contained in the tar archives as well. Alternatively
you could manually copy the cciss.o files into the appropriate
/lib/modules subdirectory and manually modify your boot loader's
configuration if you so choose.)

Or, there are several environment variables which will suppress
various aspects of the RPM package's post (un)installation behavior:

NOINITINSTALL=y
Suppress making multiple initrd images for all base and errata
kernels. Create the initrd image for the currently active
kernel only.

SUPPRESSINITRD=y
Suppress initrd, LILO, and GRUB modifications.

DRYRUN=y
Perform only a "dry run" of the post-install steps. Copying of
driver object modules, and LILO, GRUB, and initrd configuration
changes will be "simulated" but not actually performed. (Note,
normally a "dry run" is done automatically once prior to
performing the real installation as a safeguard. If the "dry run"
fails, the real installation is not performed. Setting the "DRYRUN"
variable simply makes the installation halt after the "dry run"
even if the "dry run" succeeds.) Note also that the RPM is still
considered to be "installed" and "rpm" will refuse to install it
again, so if you use the "DRYRUN" option, you'll have to uninstall
the RPM via:

rpm -e --noscripts cpq_cciss-2.4.60-13

before reinstalling it without the "DRYRUN" variable set to enable
LILO, GRUB and initrd modifications to proceed. Alternately, you
could manually unpack the gzipped tar file and use the "install"
script to install the driver object modules, make a new initrd
image and change your boot loader configuration, or manually
perform those steps yourself.

NOTE: The /etc/modules.conf file will not be updated. Most of the time,
this will not be a problem, as the drivers installed by this RPM will
be updates to previously installed drivers. If, for whatever reason,
the RPM installs a driver for which there is not already an entry in
/etc/modules.conf, then the driver will not be loaded automatically at
boot time, and it will be necessary to use insmod manually, or modify
/etc/modules.conf manually. If /etc/modules.conf must be modified
manually to add an entry for the driver, then a new initrd image
must also be created, since mkinitrd uses modules.conf to determine
what to put into the initrd image.
NOTE: The above holds true to SUSE distributions as well, except that
the file that contains the list of modules to be placed in the initrd
is /etc/sysconfig/kernel rather than /etc/modules.conf. Likewise the
initrd command is "mk_initrd" rather than "mkinitrd".


You can run "rpm -q --scripts cpq_cciss" (if installed) or "rpm -q --scripts -p " (if not installed) to see whar kernel versions are supported.

-Ross

[Tux_1]

If you replace UUID with this value /dev/cciss/c0d0p3 in GRUB doesn't work , why ?