Known Issues

System appears to hang during installation or boot

If the system appears to hang during boot or installation, messages similar to the following are displayed:

Loading initrd initrd-2.6.9-68.EL.img...done

This can occur when the firmware setting for the primary console is not set correctly for the console device being used.

Use the EFI Boot Manager interface, that is initiated during the power on sequence, to set the system console firmware properly for the device being used by choosing one of the following:

Use of a serial console when a VGA card is present

If the Red Hat kernel detects a VGA card, it defaults to using the VGA display as system console.

To use serial console, use the EFI Boot Manager to select the serial device as the primary (or only) console device as described in the previous issue.

Applications may hang when opening Management Processor UART

The Management Processor (MP) UART on all HP Integrity servers running RHEL do not supply the carrier detect signal. This causes applications to hang when opening the UART device because no carrier detect is sent unless the O_NDELAY or O_NONBLOCK flag is set.

For example, the command echo foo > /dev/ttyS0 would hang.

Typically, this is not a problem because /dev/ttyS0 is usually used as a console, the agetty process opens it with O_NONBLOCK, and processes spawned by agetty generally inherit the already-opened device. However, there are some cases where the device must be opened again so applications will observe the hang.

Use the following procedure to resolve the problem:

Installation failure when using autopartition

The installation of RHEL AS 4.7 might fail if using the autopartition function on a server with multiple disks.

If you encounter this issue, remove all disks from the server except those on which you are installing RHEL AS 4.7. Replace the removed disks once the OS installation on the targeted disks is complete. You may then need to edit the device file name of the /boot/efi mount point in the /etc/fstab file because the device file may have changed when the disks were replaced.

Simultaneous installations of RHEL 3 and 4

If you have RHEL AS 3 installed on a disk partition on your server and want to install RHEL AS 4.7 on another, you must change the name of your boot manager entry on the RHEL AS 3 partition. This is because RHEL 3 and 4 use the same name for the boot manager entry.

Before installing RHEL AS 4.7, you must change the name of the RHEL AS 3 boot entry. The EFI Boot Manager does not have a rename function to use to rename the RHEL AS 3 entry so you have to add a new boot entry.

To add a new EFI boot entry, select the Boot Option Maintenance menu (also called Boot Configuration in new EFI Boot Managers), and then add a new boot entry for RHEL AS 3 using a unique name with the same boot parameters as the existing RHEL AS 3 entry.

Load order of drivers may be altered

If you are transitioning to RHEL AS 4.7 from either RHEL 3 or RHEL 2.1, recognize that the load order of storage and networking drivers in this OS version is the reverse of the order in RHEL AS 3 and 2.1. The numbers on the device file names for your NICs appear in the opposite order than the previous versions of RHEL.

Load of RHEL failed error

After installing the operating system, if a message similar to the following is displayed, you must take action.

Loading.: Red Hat Enterprise Linux Load of Red Hat Enterprise Linux failed: Not Found Press any key to continue

Execute the following steps from the EFI shell:

HP SAS Integrated Raid (IR) Configuration Utility May Cause Kernel Panics

The Integrated Raid (IR) Configuration Utility, cfggen, is a Linux command line utility that configures the IR functionality of the HP Serial Attached SCSI (SAS) controllers that are used in LSI-1068-based HP SAS controllers. The cfggen utility causes kernel panics when creating volumes with RHEL4U7.

Until this issue is resolved in the MPT driver, HP recommends that you use the EFI executable, cfggen.efi, utility to create volumes.

If cfggen is installed on a system running RHEL4U7, the HP Support Pack installer will remove it.

For information about using the cfggen.efi utility, see the SmartSetup Scripting Toolkit Deployment Guide:

http://www.docs.hp.com/en/5991-6250/

Possible MCA with some ELILO bootloader options

The ELILO bootloader can pass command-line options to the RHEL kernel. The max_addr= and mem= options limit the amount of memory used by the kernel.

Some versions of the kernel handle these arguments incorrectly, resulting in an MCA that crashes the system during boot. If the system boots successfully, there is no risk of a crash due to this problem.

Booting with max_addr= or mem= can be useful for debugging problems, but is not a tested feature. You can achieve similar effect by changing the configuration with the EFI dimmconfig command or physically removing DIMMs.

Intermittent FTP data truncation when using an HP Procurve 4108gl switch

If you are using an HP Procurve 4108gl switch on a heavily loaded system, you may experience intermittent data truncation with FTP transfers through the network. This can be avoided with the following steps:

nPartition ACPI setting must be correct before booting RHEL on some servers

Before booting RHELAS 3 or 4 on an nPartition on an rx7620, rx8620, or sx1000 server, your ACPI configuration value must be set to single-pci-domain. This eliminates any bus address conflicts and ensures that all I/O slots have a unique address. Although your server should have the correct setting already, confirm it before proceeding and change it if necessary using one of the following methods:

Powering off nPartitions with shutdown or poweroff commands

The default behavior on rx7620, rx7640, rx8620, or rx8640 servers is for an nPartition to be made inactive (all cells are in a boot-is-blocked state) when shutdown -h or poweroff is issued from the RHEL command line. This behavior is established with the acpiconfig disable powerdown setting, which is the default setting for the single-pci-domain ACPI configuration.

To change the default behavior and shutdown nPartitions automatically, run the acpiconfig enable softpowerdown command from the EFI shell, and then reset the nPartition to force the new ACPI configuration take effect.

When softpowerdown is enabled on an any of these servers, if only one nPartition is defined in the server, then halting the operating system powers off the server cabinet including all cells and the I/O chassis. On an rx7620 or rx8620 server with multiple nPartitions, halting the operating system from an nPartition with softpowerdown enabled causes only the resources on the local nPartition to be powered off. You can run the acpiconfig command with no arguments to verify the current setting and the softpowerdown setting; the softpowerdown information is displayed only when different from the default behavior.

To power on hardware that has been powered off, use the PE command at the MP command menu. To make an inactive nPartition active, use the MP BO command to boot the nPartition past the boot-is-blocked state.

	NOTE: On HP Integrity Superdome servers, an nPartition is always made inactive when halted from the operating system (for example, after shutdown -h is executed), and this behavior cannot be changed.

System panics resulting from Fibre Channel storage arrays directly connected to the system

System panics may occur on HP Integrity servers running RHEL when Fibre Channel storage arrays are directly connected to the system.

If you plan to connect Fibre Channel storage arrays to HP Integrity servers running RHEL, you must connect through a Fibre Channel switch. For the BL860c Server Blade, you must use either the Fibre Channel switch module (AE370A, AE371A, or AE372A) or an external Fibre Channel switch.

Machine check aborts (MCA) occurring on cell-based servers

HP mid-range and high-end (cellular) systems can be configured with varying amounts of cell local memory (logical memory ranges mapped to contiguous physical memory on the cell) or interleaved memory (logical memory ranges distributed across physical memory in all cells).

If a system does not allocate enough cell local memory (CLM) to any given cell, an MCA could occur. Should this happen, try one of the following to resolve the issue:

If the system was working previously, issue the pdt command at the EFI prompt. If there are PDT entries, consider replacing the bad memory.

32-bit emulation and other 32-bit functionality

The 32-bit emulation is not automatically installed with RHEL AS 4.7. To obtain this functionality, install the ia32el package from the RHEL AS 4 for Itanium Extras Disc.

You can install any additional compatibility binaries needed from the RHEL AS 4.7 Compatibility Layer Disc. For instructions, see the RHEL AS 4 Release Notes as described in “Installation Documentation”.

Use of the cdrecord program

The cdrecord program may hang when burning a CD-RW disk with unsupported media using certain models of CD/DVD combo and DVD+RW drives shipped in rx1600, rx1620, rx2600, rx2620, and rx4640 servers. Use only approved HP CD-RW media to prevent burn hangs. Try different brands of media if HP media is not available.

In addition, when using cdrecord to burn a disc, you may need to eject the disc and reinsert before you can mount it.