Memory Fault Tolerance - IBM System x3850 X5 Implementation Manual

Important: Never warm-boot the server while running the built-in diagnostics. Several
built-in functions that are used in the normal operation of the server are disabled during
diagnostics to get direct results from the hardware. Only a normal exit from diagnostics or a
cold boot of the server will re-enable those functions. Failure to perform this task correctly
causes the server to become unstable. To correct this problem, simply power off the server
and power it back on.

6.3.2 Memory fault tolerance

For servers with high availability requirements, using the memory mirroring or memory
sparing configuration allows the server to continue to function normally in the rare event of a
memory failure. See 2.3.6, "Reliability, availability, and serviceability (RAS) features" on
page 28 for an explanation of memory mirroring and the sparing function.
Considering that memory is a solid-state device (SSD), it is unlikely that a DIMM failure will
occur outside of the first 90 days of operation. Statistically, there is a higher risk for failure with
power supplies, copper network adapters, storage devices, or processors. High availability is
almost always a desired goal, but for true high availability, consider a cluster of host
computers using common storage that allows virtual servers to be defined and moved from
one host server to another host server.
If high availability is the most important aspect of your server (with cost and performance as
secondary concerns), memory mirroring or sparing are good features to enable.
Use the following steps to establish memory mirroring or memory sparing:
1. Boot the server into F1-Setup by pressing F1.
2. From the System Configuration panel, select System Settings  Memory. See
Figure 6-6.
Figure 6-6 The memory configuration panel in F1-Setup
Chapter 6. IBM System x3850 X5 and x3950 X5
229