Troubleshooting System problems - IBM BladeCenter (8677)

Issuing any of the Linux shutdown commands shuts down the operating system, but does not turn off the blade server. The blade server shuts down to the point where a message is displayed on the management module saying that the blade server is off, but the blade server power-on LED remains lit.

Complete the instructions provided for Red Hat Linux in "Configuring Wake on LAN in the BladeCenter unit" on page 25 of the BladeCenter Hardware Maintenance Manual to enable Wake On LAN for both onboard Broadcom integrated Ethernet controllers. This will allow the blade server to restart using Wake on LAN even if the Linux shutdown is incomplete and the blade server does not turn off. Forcing a blade server to turn off.

Complete the following steps to turn off a blade server after an incomplete Linux shutdown:

1. Manually press the power-control button on the front of the blade server (behind the control panel door) to turn off the blade server. Turning off the blade server in this way is considered an improper shutdown by the Wake on LAN feature, and the Wake on LAN feature will not restart the blade server.
2. Unlatch and slide the blade server partially out of the BladeCenter; then, reinstall it to reset the blade server power state and re-enable its previously programmed Wake on LAN capability.

Issuing any of the Linux shutdown commands shuts down the operating system but does not turn off the blade server. The blade server shuts down to the point where a message is displayed on the management module saying that the blade server is off, but the blade server power-on LED remains lit.

A blade server running SUSE Linux 8.0 will not turn off during shutdown if ACPI is not enabled. Complete the following steps to enable ACPI:

1. Type the following command:
   modprobe ospm_system
2. Type the following command to save this setting upon subsequent reboots:
   echo ./sbin/modprobe ospm_system. >> /etc/init.d/boot.local

The problem occurs after installing the second microprocessor option or after replacing a failed microprocessor in a two-way Blade server.

The processor with the lowest feature set and stepping level must be used as the Bootstrap Processor (BSP). This is Microprocessor 1 in location U66. Move the microprocessor in location U66 to location U70, and move the microprocessor in location U70 to location U66.

Before switching ownership of the media tray to another blade server, safely stop the media tray devices on the blade server that currently owns the media tray, as follows:

1. Double-click the Unplug or Eject Hardware icon in the Windows taskbar at the bottom right of the screen.
2. Select USB Floppy, and click Stop.
3. Select USB Mass Storage Device, and click Stop.
4. Click Close.

Attempting to access the mounted CD-ROM drive or diskette drive (media tray) after it has been switched to another blade server results in I/O errors, even if the media tray has been switched back.

Note: Because the BladeCenter unit uses a USB bus to communicate with the media tray devices, switching ownership of the media tray to another blade server is the same as unplugging a USB device.

• If a blade server tries to access the CD-ROM after it has been switched to another blade server, a "Media not found" error occurs.
• If a blade server is running a DOS environment, such as when updating firmware on the blade server, the firmware can be interrupted or corrupted when the media tray is switched away; you might need to call for service on the blade server.
• If a file handle was left open by switching the media tray away, the system administrator will not be able to do a clean unmount (umount command) unless the unmount is forced by umount command parameters ("lazy umount").
• If the system administrator is sharing out the CD-ROM drive for multiple users, that network share is broken.Before switching ownership of the media tray to another blade server, ensure that the CD-ROM drive and diskette drive are not mounted for the current blade server owner (check for open file handles and sharing out). If a firmware update is taking place on the blade server, do not switch the media tray to another blade server.

Neither Red Hat Linux 7.3 nor SUSE Linux version 8.0 will install locally, although Red Hat Linux might appear to install.

• If you try to install Red Hat Linux to the blade server IDE drive, Linux does not install.
• If you try to install Red Hat Linux to the SCSI drive on a blade server expansion unit, Linux appears to install but the operating system will not start properly.
• If you try to install SUSE Linux, Linux does not install.

Download the latest operating system installation instructions for your operating system from the IBM Support Web page at http://www.ibm.com/systems/support/. The necessary workaround is described in the instructions for your operating system.

The X Window does not start in a blade server if the blade server is not the current owner of the keyboard, video, and mouse (KVM). Because the blade server does not own the KVM, the X Window cannot find the mouse.

Depending on the Red Hat Linux kernel version, do one of the following:

• Red Hat Linux 7.3, kernels prior to 2.4.18-17.7.x, and Red Hat Advanced Server kernels 2.1: Add the following line to the /etc/modules.conf file.
  alias char-major-13 mousedev
• Red Hat Linux 7.3, kernel 2.4.18-17.7.x or later: Install the 2.4.18-17.7.x or later kernel errata, available from https://rhn.redhat.com/errata/rh73-errata.html.

The remote console requires a display setting of 1024x768@60Hz in the blade server operating system. The default resolution in SUSE is 1024x768, but the default refresh rate falls somewhere between 50Hz and 60Hz. The remote console does not work for a blade server running SUSE with a display refresh rate other than exactly 60Hz. The message "eServer/No video available" displays.

Other operating systems do not exhibit the problem.

Set the refresh rate in the XF86Config file to exactly 60Hz.

There are two methods.

• Method 1 (unattended network install, prevent the problem): Modify the AutoYaST control file to specify 1024x768@60Hz.
  1. Run the graphical interface to the AutoYaST control file to set the VESA video mode to 1040x768@60Hz. The graphical interface creates the AutoYaST control file.
  2. Edit the resulting AutoYaST control file to set the value for min_vsync to 60.

  During an unattended network install, the YaST program uses the AutoYaST control file to modify the XF86Config file (/etc/X11/XF86Config); these changes will cause XF86Config to set the display resolution to 1040x768 with a refresh rate of 60Hz.
  
  -- OR --
  

• Method 2 (situation has already occurred): Modify the xF86Config file.
  1. In the Monitor section of /etc/X11/XF86Config, change the value of VertRefresh to 60, as shown in these sample lines. Section "Monitor"
     Option "CalcAlgorithm" "IteratePrecisely"
     HorizSync 31-48
     Identifier "Monitor[0]"
     ModelName "AutoDetected"
     Option "DPMS"
     VendorName "AutoDetected"
     VertRefresh 60
     UseModes "Modes[0]"
     EndSection
  2. Shut down X; then, restart it. Diskette drive problems

Both Red Hat Linux and SUSE Linux see the diskette drive as /dev/sdx (where x is the last SCSI device in the line) instead of /dev/fd0. When you try to mount the drive as fd0, the Linux operating system does not recognize the device ("not a valid block device"). There are two options:

• (For SUSE Linux version 8.0)
  Modify the /etc/fstab file to include the following line:
  /dev/sdx /media/floppy auto noauto,user,sync 0 0
  where x is the letter that was assigned to the diskette drive.
• (For Red Hat Linux 7.3)
  1. Modify the /etc/fstab file to include the following line:
     /dev/fdx /mnt/floppy auto noauto,user 0 0
     where x is the letter that was assigned to the diskette drive.
  2. Create a directory for the diskette drive under /mnt by running:
     mkdir /mnt/floppy
  3. Mount the diskette drive by typing:
     mount /mnt/floppy
     Note: When running Linux, you must unmount the diskette drive before changing ownership of the media tray.

1. Replace the diskette drive
2. Replace the diskette drive signal/power cable CD-ROM drive problems

If the SUSE Linux operating system is installed remotely onto a blade server that is not the current owner of the media tray (CD-ROM drive, diskette drive, and USB port), SUSE sees the CD-ROM drive as /dev/sr0 instead of /dev/cdrom, establish a link between /dev/sr0 and /dev/cdrom as follows:

1. Enter the following command:
   rm /dev/cdrom; ln -s /dev/sr0 /dev/cdrom
2. Insert the following line in the /etc/fstab file:
   /dev/cdrom /media/cdrom auto ro,noauto,user,exec 0 0

When the CD-ROM drive is owned by blade server x, is switched to another blade server, then is switched back to blade server x, the operating system in blade server x no longer recognizes the CD-ROM drive. This happens when you have not safely stopped the drives before switching ownership of the CD-ROM drive, diskette drive, and USB port (media tray).

Note: Because the BladeCenter unit uses a USB bus to communicate with the media tray devices, switching ownership of the media tray to another blade server is the same as unplugging a USB device. Before switching ownership of the CD-ROM drive (media tray) to another blade server, safely stop the media tray devices on the blade server that currently owns the media tray, as follows:

1. Double-click the Unplug or Eject Hardware icon in the Windows taskbar at the bottom right of the screen.
2. Select USB Floppy and click Stop.
3. Select USB Mass Storage Device and click Stop.
4. Click Close.

You can now safely switch ownership of the media tray to another blade server.

1. Replace the CD-ROM drive.
2. Replace the CD-ROM interposer card.
3. Replace the CD-ROM signal cable.
4. Replace the CD-ROM power cable.

Enumeration of the Ethernet controllers in a blade server is operating-system dependent. In the blade server Configuration/Setup Utility program, the Ethernet port designated as Planar Ethernet 1 is routed to Ethernet switch module 2 and the Ethernet port designated as Planar Ethernet 2 is routed to Ethernet switch module 1. Verify the designations through your operating system settings or by testing:

1. Install only one switch module, in switch bay 1.
2. Enable only one of the Ethernet controllers on the blade server. Make note of the designation the blade server operating system has for the controller.
3. Ping an external computer on the network connected to the switch module.If you can ping the external computer, the Ethernet controller you enabled is the upper controller in the blade server and is associated with Ethernet switch 1.

When you use the Ethernet switch module Telnet interface to request the switch module to ping something, the first ping response reports a failure, although the other repetitions might report success. This occurs regardless of whether the switch module port the pinged object is connected to is internal or external, and applies to pinging blade servers but not to pinging the management module or objects connected to its external Ethernet port, such as the network management station.

To get accurate results, always specify multiple repetitions (>1) in the ping request, and ignore the first ping response from that request. See the IBM 4-Port Gb Ethernet Switch Module for BladeCenter Installation Guide for instructions on how to ping through the Telnet interface.

If the Wake-on-LAN (WOL) feature is disabled on a blade server, and the blade server is turned off, the switch module internal port link to that blade is down. This is not an error, but the graphic of the BladeCenter unit might show a blank panel instead of a blade server in that bay.

Note: You can enable or disable the WOL feature on a blade server through the management-module Web interface or through the IBM Director console. Do not rely on the BladeCenter graphic in the Ethernet switch-module firmware Web interface to determine the presence or absence of blade servers in the BladeCenter unit.

When you use the management-module Web interface to update the Ethernet switch module configuration, the management module firmware writes its settings for the switch module only to the management module NVRAM; it does not write its settings for the switch module to the switch-module NVRAM. If the switch module restarts when the management module is not able to apply the IP address it has in NVRAM for the switch module, the switch module will use whatever IP address it has in its own NVRAM. If the two IP addresses are not the same, you might not be able to manage the Ethernet switch module any more. The management module cannot apply the switch IP address from its NVRAM if:

• The management module is restarting.
• The management module has failed.
• The management module has been removed from the unit.

When you use the management-module Web interface to update the Ethernet switch module configuration, the management module firmware writes its settings for the switch module only to the management module NVRAM; it does not write its settings for the switch module to the switch module NVRAM. If the switch module restarts when the management module is not able to apply the IP address it has in NVRAM for the switch module, the switch module will use whatever IP address it has in its own NVRAM. If the two IP addresses are not the same, you might not be able to manage the Ethernet switch module any more. The management module cannot apply the switch IP address from its NVRAM if:

• The management module is restarting.
• The management module has failed.
• The management module has been removed from the unit.

Although the keyboard attached to the BladeCenter unit is a PS/2-style keyboard, communication with it is through a USB bus. When you are running an operating system that does not have USB drivers, such as in the following instances, the keyboard responds very slowly.

• Run the blade server integrated diagnostics.
• Run a BIOS update diskette on a blade server.
• Update the diagnostics on a blade server.
• Run the Broadcom firmware CD for a blade server.

When you are redirecting the server console (remote console function) of a blade server that is running Microsoft Windows 2000 or Windows XP and using the Sun Java plug-in (Java Virtual Machine), the remote console can have keyboard entry problems. Use the Microsoft Java Virtual Machine (JVM) or Java Runtime Environment (JRE) on the blade server instead of using the Sun Java Virtual Machine. The Microsoft JVM comes with the Windows XP Service Pack

You can obtain the Microsoft JVM for Windows 2000 from the Microsoft corporation. If you are using the Internet Explorer browser, version 6.0 or later, to log into the management module and use the remote control function, you must also adjust the browser settings:

When the BladeCenter management module is restarted, use of the media tray (CD-ROM drive, diskette drive, and USB port) is lost temporarily. If you or a failure condition initiates a management module restart while I/O activity is taking place on the media tray, the disruption can interrupt reads to the CD-ROM drive or lose data being written to a diskette.

Note: You can restart the management module through the Web interface to the management module or from a network management station such as the IBM Director console. Some failures on the BladeCenter unit can result in the management module restarting automatically. Make sure there is no I/O activity on the media tray before you restart the management module.

Verify that:

1. The system power cord is plugged into the BladeCenter power module and a working 220-volt PDU or electrical outlet.
2. The monitor cables are connected properly.
3. The monitor is turned on and the brightness and contrast controls are adjusted correctly.
4. If you have verified these items and the screen remains blank, replace:
   1. Monitor
   2. Management module

If the monitor self-tests show the monitor is working properly, consider the location the monitor. Magnetic fields around other devices (such as transformers, appliances, fluorescent lights, and other monitors) can cause screen jitter or wavy, unreadable, rolling, or distorted screen images. If this happens, turn off the monitor. (Moving a color monitor while it is turned on might cause screen discoloration.) Then move the device and the monitor at least 305mm (12in) apart. Turn on the monitor.

NOTES:

1. To prevent diskette drive read/write errors, be sure the distance between monitors and diskette drives is at least 76mm (3in.).
2. Non-IBM monitor cables might cause unpredictable problems.
3. An enhanced monitor cable with additional shielding is available for the 9521 and 9527 monitors. For information about the enhanced monitor cable, contact your IBM reseller or IBM marketing representative.If the problem remains, replace the monitor.

Verify that:

1. The power cables are properly connected to the power modules in the BladeCenter unit.
2. The 220-volt PDU functions properly.
3. The LEDs on the power module are on.
4. If you just installed an option, remove it, and restart the BladeCenter unit. If the BladeCenter unit now turns on, you might have installed more options than the power module supports. You might need to install a power module in power bay 3 or 4.If the problem remains, go to "Undetermined problems" on page 90 of the BladeCenter Hardware Maintenance Manual.

Verify that:

• The option is designed for the BladeCenter unit. See the "Server Support" flowchart in the
• BladeCenter Hardware Maintenance Manual for information about obtaining Server Proven compatibility information from the World Wide Web.
• You followed the installation instructions that came with the option.
• The option is installed correctly.
• You have not loosened any other installed options or cables. If the problem remains, replace the option.