|C H A P T E R 2|
This chapter contains the following sections:
To install the Linux operating system onto a blade, you must first build a PXE boot installation environment.
The software required to build a PXE boot installation environment is available on the CD supplied with the blade.
To install Linux on B100x or B200x server blades you will need the following:
Refer to Chapter 4 of the Sun Fire B100x and B200x Server Blade Installation and Setup Guide for information on how to perform a PXE boot installation.
To run Red Hat Enterprise Linux version 3.0 or SuSE Linux Enterprise Server 8, service pack 3 on a B200x server blade, the blade must have BIOS version 1.1.32 installed. This version of the BIOS is available from the following website:
You can upgrade the BIOS on the blade using the biosupdate utility. This utility loads a device driver called mtdbios, performs the BIOS update using the slflash utility, and then unloads the mtdbios driver.
The biosupdate utility is installed on the server blade as part of the PXE boot installation process. For information on the PXE boot installation, see Chapter 4 of the Sun Fire B100x and B200x Server Blade Installation and Setup Guide (available on the CD supplied with the blade).
Note - If the BIOS upgrade fails, a failure message is displayed on the screen and in /var/log/messages. If this problem occurs do not reset or power off the blade. Contact your Sun Beta support manager for advise.
1. Log into the blade for which you want to update the BIOS.
At the SC prompt, type:
where n is the number of the slot containing the blade.
2. Check the version of the BIOS currently running on the blade, to establish whether the upgrade is necessary:
3. Copy the BIOS image from the beta website to a known location on the blade.
4. Run the biosupdate command:
The blade prompt returns when the update is complete.
5. When the update is complete, reboot the blade:
You can check the BIOS version when you restart the blade.
The following problems have been observed on both B100x and B200x server blades:
The Advanced Server kernel (version 2.4.9-e.3) does not support layer 2 VLANs. This means that the sunconfig utility is not installed on blades running Red Hat Advanced Server 2.1. If you are using blades with Red Hat Advanced Server 2.1 installed, you must configure the switch to use only untagged VLANs.
If you require VLAN support, you must install Red Hat Enterprise Linux 3.0 or SuSE Linux Enterprise Server 8 service pack 3.
Spurious interrupt IRQ7 messages may appear in /var/log/messages. These messages can be ignored.
The following problems have been observed on B200x server blades only:
When the BSC driver is loaded, five extraneous LED files appear in the /proc/bsc directory. The LED status for these files is reported as "not present" and they can be ignored.
If you attempt to boot Enterprise Linux 3.0 update 1 with APM enabled, the system will panic. This is due to expectations made by the APM subsystem, which is considered unsafe to run on a multi-processor machine. Without enabling APM (or ACPI), the /sbin/poweroff command will not power off the blade.
This problem has been worked around using the bsc driver. The bsc driver notifies the hardware of the intent to power off, and the hardware powers off the blade 10 seconds later. If the bsc driver is not loaded at power-off time, the system will fail to power off.
The CMOS footprint may change between revisions of the BIOS. If this is the case, and the CMOS is not reset to its default values during the blade reboot following a BIOS update, the CMOS configuration may become corrupted and result in repeated network boots.
To avoid this problem, after updating the BIOS on a B200x blade, reset all CMOS settings to their default values when you next reboot the blade. You can do this by typing the bootmode reset_nvram sn command at the SC prompt (where n is the number of the slot containing the blade), or by entering the BIOS setup menu and loading the BIOS defaults.
When new 1 Gbyte DIMMs are inserted in DIMM slots on the B200x blade, the ECC driver incorrectly identifies the memory as 2x512Mbyte DIMMs.
This is a cosmetic issue, and can be ignored.
When a blade is booted without an SSC present, the network interfaces normally attached to the missing switch will not be operational after the SSC is re-inserted. This problem is fixed when the network interface is brought down and up again.
The following problem has been observed on B100x server blades only:
When you use the reset command from the SC prompt, the blade updates the Real Time Cock (RTC) from the system controller through the BSC. The RTC should be set only when the blade is powered on.
The Operating System hangs when the bsc driver attempts to allocate IRQs 9 and 5. This is due to an incorrect ACPI table entry for the bsc hardware.
You can avoid this problem by booting the kernel using the pci=noacpi argument.
This is a cosmetic issue and may be ignored.