Bug 452056

Summary: MBR gets installed in the wrong place on HP DL585
Product: Red Hat Enterprise Linux 5 Reporter: Jeffrey Needham <jneedham>
Component: anacondaAssignee: Anaconda Maintenance Team <anaconda-maint-list>
Status: CLOSED NOTABUG QA Contact: Release Test Team <release-test-team-automation>
Severity: medium Docs Contact:
Priority: medium    
Version: 5.2CC: green, hdegoede, james.brown
Target Milestone: rc   
Target Release: ---   
Hardware: x86_64   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2009-11-30 10:36:23 UTC Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:

Description Jeffrey Needham 2008-06-19 01:44:37 UTC 
Description of problem:
During a fresh install or an update install, the MBR gets installed in the wrong
place.

Version-Release number of selected component (if applicable):


How reproducible:
Try to install or update install RHEL5.2 (from RHEL5.1 in this case) on a HP
Proliant that has both SW RAID and FCP LUNs attached. The MBR ends up on FCP LUN
0 (/dev/sda) and not on the SW RAID lun (/dev/cciss/c0d0)

Steps to Reproduce:
1.Any proliant with an SW RAID and an FCP HBA
2. Register the HBA's WWID so that LUNS appear upon boot (/dev/sda, /dev/sdb etc)
3. Do a initial or update install
  
Actual results:
The installer insists in the initial install case on putting the MBR on
/dev/sda, not on /dev/cciss/c0d0.

In the update install case, it automatically updates it to the wrong place.


Expected results:
In the UPDATE INSTALL case, /dev/cciss/c0d0 should continue to hold the MBR.  In
the INITIAL INSTALL case, /dev/cciss/c0d0 should be as the boot device.

Additional info:
FCP HBA was a QLA 2312.
Luns supplied by a Netapp 960
SW RAID lun consisted of 4 36GB U320 drives configured as 1 RAID5 volume
The DL585 had 64GB of memory and 4 AMD 250 CPUs.
The BIOS IPL order had the PCI Embedded RAID appear before the Qlogic HBA
Comment 1 Anthony Green 2009-05-27 15:10:13 UTC 
I've seen this as well on a Proliant BL680C.  The storage configuration is the same (cciss and SAN luns).

Work around is as follows: 

* Review partitions during install
* Select "Configure advanced boot loader options".
* From this screen you can change the Driver Order to place the local disk at the top of the list, and select c0d0 as the target for the MBR installation.

Anaconda should be smarter here, as we've already told it not to touch /dev/sd* on a previous screen.
Comment 3 Hans de Goede 2009-11-30 10:36:23 UTC 
There are 2 issues here:

1) anaconda can not reliable detect from which drive the BIOS will boot the system
   this is always true, but esp. so in the case when booting from an add in card
   with a separate option ROM. As this makes detecting  which drive the BIOS is
   set to boot from even harder. So we cannot automatically get this right.

   Another example where we fail to get this right is when installing from a
   usb stick, because then usually we do manage to find out which disk the
   system is booted from through the EDD BIOS extension, but the BIOS then tells
   us the system is booted from the usb stick, which is correct, but usually
   not the place where the user wants to install the bootloader.

   So there will always be cases where one needs to use the advanced bootloader
   options as described in comment #1.

2) There are several scenario's where users want to not use a disk for 
   installation, yet still put the bootloader there, so we cannot assume that
   if a disk is not used for installation, we should not put the bootloader there.