[wplug] Install Question

[Bryan J. Smith]

On Wed, 2007-07-25 at 11:19 -0400, Michael H. Semcheski wrote:
> I'm setting up Scientific Linux (for all intents and purposes, read
> that CentOS) on a brand new machine.  This machine has a 3ware RAID
> card.

Which 3Ware/AMCC card?

The 64-bit ASIC-SRAM designs are 7000/8000 series (have no idea about
the microcontroller core).  They use the 3w-xxxx driver and have
firmware that has been long stablized (7.7.1 -- circa 2004+).  And since
they are SRAM, they don't need a BBU (battery backup unit) as a few caps
are enough to sold the charge necessary to keep SRAM data.

The 64-bit ASIC-SRAM+DRAM are the 9500S.  This was really not an ideal
design, since the ASIC was insufficient for DRAM buffering.  This uses
the 3w-9xxx driver.

The 9500S was quickly replaced by 9550SX and latter products after the
purchase of 3Ware by AMCC, the official embedded PPC400 series licensee.
The latest rev is pretty idea in the performance department, and the
PPC400 series does very well against Intel XScale IOP330/340 series.
This also uses the 3w-9xxx driver.

It's _very_important_ with any "true hardware" RAID card to keep the
firmware up-to-date and "in-sync" with the driver.  Otherwise the kernel
may communicate to the on-board microcontroller incorrectly.  Luckily
3Ware/AMCC has been bundling the firmware object code in the 3w-9xxx
driver in the stock kernel for quite awhile now, so that shouldn't be an
issue.

[ SIDE NOTE:  On the elementary (although hardly extensive) differences
between non-blocking ASIC+SRAM "caching" and uC+DRAM "buffering"
controllers (as well as "software/OS" versus "[software] Fake/Free RAID,
FRAID, driver"), see my (now aged) "Dissecting ATA RAID Options" article
in Sys Admin 2004 April:  http://www.samag.com/documents/sam0404c/ ]

> The RAID5 configuration is 4x750.  (So I'm installing onto a
> roughly 2TB disk, using the default partitioning.)
> Anaconda does fine.  No errors.  After the disk is formated and the
> packages are copied over from the CD's, it asks me to reboot.  I
> reboot, and after the BIOS and RAID are initialized, instead of seeing
> grub, I just see a blank text screen with a blinking cursor in the top
> left corner.

What events do you see in the 3Ware EEPROM?  You should be able to get
into the 3Ware firmware at POST (power on self-test) by hitting "Alt-3"
at the appropriate moment.

Remember, you have a true hardware RAID controller with full, on-board
intelligence.  It's like having a standalone chassis, just without the
chassis and separate power/storage bus (using PCI-X/PCIe instead).  It
logs everything.

If you don't have any troubling events, then the installer failed to
install the GRUB bootloader correctly.  A BIOS Int13h Disk Services disk
mapping (0x80 = "Drive C:/Disk 1") to Linux device (/dev/sda) may be the
issue.  On Fedora-based distros, this is set in the device
"/boot/grub/device.map" such as follows ...
  (hd0)     /dev/sda

Do you have _any_ other storage devices in the system that may upset
this mapping?  An ATA (IDE) or SATA device perhaps?  Or possibly a USB
storage device that got loaded and mapped to /dev/sda during install?

> Any ideas what's going on?

Plenty.  I have troubleshooted over 1,000 boot-time issues in the last
twelve (12) years with Linux -- from BIOS mapping, initrd, multipathing,
etc... on ATA, SCSI, SAS, FC and other local, subsystem and SAN
technologies.  It can be a mess if you don't know all the little things
to consider -- they all have to "match up" (map) correctly.

God I miss OpenBoot, SRM and other "intelligent" UNIX firmware.  In
fact, nothing amazes me more than when some Microsoft cronie acts like
PXEBoot and Installation Service are "new concepts."  ;)


-- 
Bryan J. Smith         Professional, Technical Annoyance
mailto:b.j.smith at ieee.org   http://thebs413.blogspot.com
--------------------------------------------------------
        Fission Power:  An Inconvenient Solution