IBM X366 Service And Support Considerations

xSeries 366

Service and Support Considerations

Rev 2.7.1

March 2, 2006

IBM CONFIDENTIAL

IBM CONFIDENTIAL

1

Overview

This document is intended to assist the Service and Support teams in some potential
problems they may likely encounter on the x366.

IBM CONFIDENTIAL

2

I. Mechanical Concerns

A. Microprocessor tray removal

When removing the microprocessor tray through the front of the server, the memory
cards or planar will be damaged unless the memory cards are removed first. Please note
that all Fans must be removed as well before CPU tray can be removed. Also during
removal of the tray, the microprocessor heat sinks may hit or interfere with the top of the
chassis making removal difficult. To help alleviate the heat sink chassis interference, lift
up on the top of the chassis while removing the tray. A recommended lift point is via the
drive bay. Please see http://www-307.ibm.com/pc/support/site.wss/migr-60199.html for
additional information on tray removal.

IBM CONFIDENTIAL

3

B. SAS Hard Disk Backplane

Connectors on both the SAS hard disk drives and SAS backplane can become damaged if
the SAS backplane is not seated properly which can keep the Hard Disk Drives from
aligning properly on the backplane. Insure the SAS backplane is fully seated properly
before installing the hard disks. Three tabs on the backplane assembly must fully seat
into appropriate slots in the chassis.

C. x366 Capacitors: Shearing Off Elpida Memory Dimms

Warning:

The system may have intermittent memory failures if the DIMM decoupling capacitors
are broken. PC3200 DIMMs have a characteristic physical design where decoupling
capacitors are placed near the connector edge of the DIMM. The placement places the
capacitors in a position where the decoupling capacitors can easily be sheared off.

This is not a problem when one is careful to place the DIMM in the connector. A problem
may arise if the unwary hand places the DIMM in between the connectors. The capacitors
are small and will easily break off if the DIMM is placed in between the connectors.

Correct Action:

DIMMs must be carefully placed in the connector by aligning the DIMMs inside the
connector card guide and aligning the card so the connector key and the connector match.
Please see Memory Handling Guide at

http://www.samsung.com/Products/Semiconductor/Support/Label_CodeInfo/memory_m
odule_handling_guide.pdf . The guide on Samsung’s website applies to all memory

manufacturers and is the method approved by IBM.

DDR2 memory DIMMs cannot be handled in the same manner as the prior generation
DDR1 memory DIMMs. DDR1 DIMMs are more tolerant to mishandling. Improper
DDR2 DIMM handling can result in mechanical damage to the components on the
DIMMs such as cracked solder joints and solder joint separation.

D. Unable to disengage left front Bezel Cover Latch

When harddrives are installed in the two left most slots, it may be difficult and/or
impossible to press the left front bezel latch to remove the front cover. The solution is to
power the machine off and remove the left two drives, then remove bezel by pressing the
left and right bezel release buttons.

IBM CONFIDENTIAL

4

E. Power protection cap falls off

Under normal conditions, the protection cap will prevent inadvertent actuation of the
power switch. There have been reports where the protection caps will not remain
attached to the bezel. The power button protection cap is now being heat-stake attached.
Please contact the IBM support to obtain a replacement.

F. Force Power On jumper on I/O board s labeled incorrectly in PD Guide

The Force Power On jumper on the I/O board is labeled incorrectly in the PD Guide. The
layout in the PD Guide show pins 1 2 3 from left to right but actually they are 3 2 1. Pins
1 and 2 should be jumpered to force power on.

Please note that as of right now, the Force Power On Jumper does not by-pass the front
panel switch. Once the jumper is moved, the fans will spin up and the server will appear
to power on and hang at CP 0b. with nothing appearing on the screen. The CR11 led on
the PCI-X board will also be out. You must press the front panel power button in order
for the server to fully power on. Once the button is depressed, the CP codes will scroll
through as normal, the PCI-X CR11 led will come on solid, and shortly after the server
will post text on the screen as normal.

G. Bent Pins on memory board and possible F2 Checkpoint hang

If the server hangs during post at checkpoint F2 listed on Super I/O card, then check the
memory card/s for possible bent pins. Replace the card if found to have bent pins.

H. USB Floppy drive does not ship with the server

A floppy drive is not included with this server. If there are updates that can only be done
through bootable diskette, then the customer must provide his own USB floppy drive.

I. Problems Accessing CDROM

If a customer is having problems accessing CDROM’s during normal operation, then have them
update the CPLD code to version 1.01. CPLD code can be obtained at http://www-

307.ibm.com/pc/support/site.wss/ MIGR-60630.html .

IBM CONFIDENTIAL

5

II. Lightpath

A. System Information Light comes on when BMC System Event Log Full

The baseboard management controller (BMC) system event log (SEL) can have a
tendency to fill up quickly with many informational type messages. Once the BMC log is
70% full (approx 358 entries), the system bang light is illuminated. In environments
where A/C power is cycled often, the tendency for to log to fill quickly is particularly a
problem. Flashing the BMC firmware will also add many informational type messages
to the SEL.

To clear the Bang Light for a full SEL, Restart the server and press F1 when prompted to start the
Configuration/Setup Utility program.
a. Select Advanced Setup;
b. Select Baseboard management controller (BMC) settings;
c. Select BMC System Event Log
d. Select Clear error logs to clear the BMC system event log.

They can also be cleared online using the SMBridge command line tools. Please refer to

http://www-307.ibm.com/pc/support/site.wss/ MIGR-57729.html for additional

information on SMBridge.

B. PCI slot Attention and Power LED’s reversed

In other systems, when viewing the PCI slots from the rear of the system, a PCI error
condition can be seen because the amber “attention” LED is light-piped to the rear of the
system. On the x366 however, the PCI slot Attention and Power LED’s are reversed so
the green “power” LED is seen on any populated slot when the system is powered on. To
see whether there is a PCI error, first check the LightPath panel. If the “PCI” LED is lit,
then remove the top cover and view the individual slot Attention LED’s to determine
which slot is affected.

IBM CONFIDENTIAL

6

C. LED indicators for a faulty VRM

For the pluggable VRMs, initial BMC firmware levels will illuminate the VRM, CPU
and PCI board LEDs during a VRM fault. Newer code versions will illuminate only the
VRM and CPU. Regardless of the BMC firmware level, sensor circuitry is unable to
tell if a VRM fault is the result of a CPU or VRM failure. Since VRMs are less
expensive, a VRM should be replaced, prior to replacing the CPU.

Add the level of BMC code that has the fix.
SP LED blinks yellow in normal operation or good condition.

D. Dim LED’s on front panel indicator:

Power off the system, remove the white flex Op Panel cable from the I/O board, and
inspect the metal contact for damage. If the metal contacts are lifted from the cable
substrate, or if there are signs of the metal being scraped, the cable should be replaced.
The plastic contact separators on the I/O board connector (where the flex cable plugs in)
are sharp and can scrape the metal contacts on either end of the cable when the flex cable
is inserted. Please see http://www-307.ibm.com/pc/support/site.wss/ MIGR-60141.html
for further information.

In addition, this symptom may result from a blown fuse on the IO Adapter, often as a
result of problems with the flex cable mentioned above.

IBM CONFIDENTIAL

7

E. Disabling onboard Broadcom Nics result in Information Light coming on:

The System Info (!) LED on the Operator Information Panel & Log LED on the
Lightpath Panel will come on at POST and stay on if the on-board NIC is disabled in
BIOS Setup. This is not a hardware problem. The "Planar/PCI Device is disabled or not
responding" message will show in the System event Log (SEL). This message is the
cause of the LED's coming on. In the upcoming BIOS release this message will be
removed. If you re-enable the on-board NIC, or clear the log, the System Info (!) & Log
LEDs should go out if your logs are less than 70% full. What actually happens is the
LEDs stay lit until you AC Power Cycle the system.

III. Power Supply Considerations

A single x366 power supply is rated a 1300 Watts when used with a 220 V ac source, and
is capable of powering a fully loaded server without redundancy. With 110 V ac, this
power rating falls to 650 Watts for the same single power supply.

A. System configuration

A single 110V ac power supply can support a maximum configuration of 1
microprocessor, 2 PCI-X adapters, 3 hard disk drives, and 4 DIMMs. If you exceed this
configuration you must either install a second power supply or use 220V ac input.

If two power supplies are installed on 110V, the server will not support a full
configuration (700-800W) with redundancy. 220V must be used in order to maintain full
redundancy with a full configuration.

B. Fuel Gauge

The NONRED LED will be illuminated when the server is operating with non-redundant
power. If a power supply or its ac power source fails, the system will move to an over
spec state.
The Over spec LED will be illuminated when there is insufficient power to power the
system. When in an over spec state, an optional power supply can be added if only one
power supply is installed. Another alternative to fixing the over spec condition is to use
a 220V ac input source.

C. Power Supply Failure and Replacement

It is advisable to check the System Event Log (SEL) for other error conditions prior to
replacing a suspected bad power supply. As an example, a shorted circuit board would
likely show improper voltage levels in the SEL due to current limiting circuit protection.

IBM CONFIDENTIAL

8