Reinserting The Speed Burst Card For Extended Outages; Effects Of Power Loss On Hx5 2-Node Or Max5 Configurations - IBM System x3850 X5 Implementation Manual

System event log
This log contains POST and system management interrupt (SMI) events and all events that
are generated by the Baseboard Management Controller (BMC) that is embedded in the IMM.
You can view the system event log through the UEFI utility by pressing F1 at system start-up
and selecting System Event Logs  System Event Log. Also, you can view the system
event log through the Dynamic System Analysis (DSA) program (as the IPMI event log).
POST event log
This log contains the three most recent error codes and messages that were generated
during POST. You can view the POST event log through UEFI by pressing F1 at system
start-up and navigating to System Event Logs  POST Event Viewer.

8.8.2 Reinserting the Speed Burst card for extended outages

In the unlikely scenario that a 4-socket HX5 configuration might be reduced to a 2-socket
configuration due to a hardware failure, you can remove the IBM HX5 2-node scalability kit
and connect the IBM HX5 1-node Speed Burst kit to increase the processor performance.
This step is not mandatory for the remaining system to continue to operate. However, connect
the IBM HX5 1-node Speed Burst kit can boost CPU performance temporarily for CPU-bound
applications by doubling the QPI links between the CPUs on the remaining HX5.
This same solution also applies if you have HX5 with MAX5 attached and the MAX5 fails
completely.
MAX5: In the unlikely event that a MAX5 fails completely, you must detach the MAX5 from
the HX5 by removing the IBM HX5 MAX5 1-Node Scalability kit to allow the HX5 to resume
normal operations.

8.8.3 Effects of power loss on HX5 2-node or MAX5 configurations

BladeCenter (BC) H by design is split into two power domains to provide redundancy to the
chassis. From a blade bay perspective, power domain 1 supplies power to blades 1 - 7 and
power domain 2 supplies power to blades 8 - 14.
If you install the primary blade server of a scalable blade complex in blade server bay 7 of a
BC H Type 8852 chassis, the secondary blade or MAX5 is installed in bay 8. This
configuration causes the blades to be split between two separate power domains. The
following situations can occur if there is a power loss to either power domain, depending on
how the scalable blade complex is implemented.
You might see the following effects of split power domains for HX5 4-socket configurations:
A loss of power to power domain 1 results in both blade servers of a scalable partition
going down. The remaining primary node in bay 7 that still has power to it might be
powered on as a single complex system. You might receive errors on reboot stating that a
non-maskable interrupt (NMI) error occurred or that there are scalability problems. This
behavior is expected due to losing the power to the processors abruptly and the second
node being still physically connected via the scalability card but unavailable.
If the scalable blade complex is implemented in stand-alone mode, a loss of power to
power domain 2 results in only the secondary blade server in bay 8 going down. The
primary blade remains powered on but still generates an error.
444
IBM eX5 Implementation Guide