IBM H Series Installation And Service Manual page 338

Configuring the High Availability Solution System With No Single Points
of Failure
Redundant adapters and mirrored disks are the only way to guarantee redundancy in
the High Availability Solution serial disk system. In this configuration, no single hard-
ware component failure can cause the serial disk system to be unavailable. If single
disk adapters are used with mirrored disks, then AIX and HACMP error notification
methods must be developed to:
1. Identify all critical adapter errors
2. Escalate a critical adapter error to an HACMP failover
The adapter outage in a single adapter configuration will cause a brief application
outage while HACMP failover transfers control of the disks and application to a
backup node.
It is strongly recommended that both disks and adapters are mirrored if possible.
This minimizes both HACMP configuration and application disruption should a disk
adapter fail.
The following table describes outages and their impacts for the minimum High Avail-
ability Solution configuration with redundant SSA adapters (2 Ethernet, 2 SSA
adapters, mirrored disk subsystem, HACMP):
Hardware Failure Failure Behavior
Description If HACMP Is Not
Configured
Node Outage or Node unavailable
AIX crash
Ethernet adapter Access to node
failure through Ethernet
lost; error log
entry.
SSA adapter None seen; error
failure log entry
Power supply or Node/disks supply
cord failure feeds made una-
vailable
A-2
RS/6000 Enterprise Server Model H Series Installation and Service Guide
Extra Work Recovery Action and
Required To Behavior If HACMP Is Con-
Provide Recovery figured
Action (in addi-
tion to normal
HACMP config-
uration)
None HACMP failover; application(s)
unavailable for brief time
during failover
None HACMP swap adapter event
moves IP address to spare
adapter. Node Ethernet IP
address unavailable for an
extremely brief period as
address is swapped.
None N/A (no failover)
None N/A (no failover)