IBM x3850 X5, Руководство по внедрению

Страница: 255 / 580

Chapter 6. IBM System x3850 X5 and x3950 X5
235
After you install and configure the MAX5 properly, a successful link between the server and
the MAX5 can be confirmed when both units power on and off as one complete unit and the
memory in the MAX5 can be seen by the server. You also see the following message during
POST:
“System initializing memory with MAX5 memory scaling”
6.5 Forming a 2-node x3850 X5 complex
The x3850 X5 server is the easiest server to scale of any of the predecessors of this class of
server. When all of the prerequisites of scaling to X3850 X5 servers have been met and the
servers are scaled, you can apply all of the updates that are provided by IBM for the
stand-alone chassis to the scale chassis as if it were a single unit. The only exception to this
rule is the replacement of either the I/O shuttle or the processor board on either node.
For rack and cable installation instructions for scaling two x3850 X5 servers, see the product
publication Installation Instructions for the IBM eX5 MAX5 to x3850 X5 and x3950 X5 QPI
Cable Kit and IBM eX5 MAX5 2-Node EXA Scalability Kit, which is available at this website:
http://ibm.com/support/entry/portal/docdisplay?lndocid=MIGR-5084861
6.5.1 Firmware requirements
Before you attempt to scale to x3850 X5, the firmware of the two servers needs to match to a
code level that meets or exceeds the minimum firmware levels that are shown in Table 6-4.
You must apply the necessary firmware to get the two servers to matched levels while the two
servers are not connected by QPI cables.
Table 6-4 Minimum firmware levels to support two scaled x3850 X5 servers
You might experience these situations if code levels do not match:
򐂰 Severe FPGA mismatch: Neither node can power on. The hardware event log, accessed
through the IMM management port, reports an FPGA mismatch.
򐂰 Minor FPGA mismatch: The system powers on, but it reports an FPGA mismatch during
POST, prior to the F1 Splash panel, and in the hardware event log of the IMM. Expect
spontaneous reboots, memory failures, processor failures, QPI link failures, and machine
check errors.
򐂰
Severe IMM mismatch: The secondary node does not power up. The IMM of the primary
node is unable to access events occurring in the secondary node.
򐂰 Minor IMM mismatch: The primary IMM is unable to collect all of the hardware information
from the secondary node. A newer version of the IMM prevents both nodes from powering
up.
򐂰
UEFI mismatch: When two scaled x3850 X5 servers are powered up, the IMM loads the
UEFI from only the primary node. The UEFI of the secondary node is never used. The
firmware level of the primary node becomes the running firmware level of both nodes.
Type Version Build
IMM 1.14 YUOO73K
UEFI 1.23 G0E122D
FPGA 1.01 G0UD29C