Reserved Sb - Fujitsu PRIMEQUEST 2000 Series Administration Manual

CHAPTER 3 Component Configuration and Replacement (Add, Remove)
3.2 High availability configuration
- Error of IO and memory cannot be detected and may cause reset during SB hot add and SB hot remove.
- Time required to complete SB hot add depends on mounted memory size, the load of CPU, the load of IO
and so on. Estimate how long time SB hot add takes by performing DR operation in advance when
system integration.
For example, time required for SB hot add is about 60 minutes if to the partition which has an SB with 384
GB and an IOU, you add an SB with the same memory size as that of the partition. Its value is measured
in the partition without any other loads.
- During SB hot add or hot remove, at least one logical CPU is used by 100 % for hot add or hot remove
process.
- After SB hot add, reboot from the operating system and reset by MMB Web-UI may operate as Power
Cycle only once.
- In RHEL7, kdump by NMI cannot be performed during "Offlining removed Memory/CPU" phase of SB hot
remove.
- While DR is performed in RHEL7, call trace may be output together with message of "WARNING: at
arch/x86/kernel/cpu/perf_event_intel_uncore.c:XXXX uncore_change_context+0xYYYY/0xZZZZ()"
(XXXX: decimal number, YYYY, ZZZZ: hexadecimal number). That is no problem.
- In RHEL7/SLES12, Turbo Boost function of CPUs which are added by SB hot add may not be used.
- In PRIMEQUEST 2400E3/2800E3/2400E2/2800E2, do not power off the partition during SB hot remove.
- "nmi_watchdog=0" must be specified to kernel option.

3.2.4 Reserved SB

In the Reserved SB function, a spare SB is mounted in the chassis in advance, the faulty SB is automatically
disconnected, the spare SB is incorporated and the partition is restarted. The spare SB intended for switching
when there is a fault is called the Reserved SB. PRIMEQUEST
2400E3/2800E3/2400E2/2800E2/2400E/2800E support the Reserved SB function.
Using a Reserved SB function has the following advantages when a hardware error occurs in the SB.
- Quick recovery is possible without reduction in SB resources.
- In a partition with one SB, recovery is possible even if the SB fails (SB degradation)
In addition, in PRIMEQUEST 2000 series, SB in a running partition can be also specified as a Reserved SB.
A Reserved SB can be used very efficiently by using this function.
An SB in a running partition can be also used as a Reserved SB. In this case, the SB will be referred to as
Active Reserved SB.
The example below shows the operation where the SB also assigned as Reserved SB is used for test
purporses..
If there is an SB failure in a partition of the production system, the firmware issues a shutdown instruction to
the partition of the test system. After the shutdown is completed, the partition is incorporated as the SB of the
production system. However, this setting is applicable only if the shutdown time of the test system can be
permitted.
FIGURE 3.19 Example of operation of a partition of a test system using the SB as a Reserved SB
(1)
(2)
(3)
No.
Production system
Test system
Partition #1 is shutdown
Description
58
CA92344-0537-08