Cluster I/O Protocols (CIP) Configuration and Management Manual (H06.16+, J06.05+)

Table 3 CLIM-to-CLIM Failover (continued)
Resulting Failover ActionProbable CausesFailure Definition
bonded interface for more than link
pulse time
No failover, but interface resources
using the interface become unavailable.
• CLIMCMD ifstop command
CLIM-initiated interface down on a
data LAN interface
No action.Failure indication, loss of link pulse,
or CLIM-initiated interface down on
maintenance interface
• NIC hardware failure
• Cable disconnected
• Switch or network down
• CLIMCMD ifstop command
Initiate CLIM or interface failover as
directed.
• SCF SWITCH CLIM command
Operator initiated failover
No failover, but the CLIM object goes
to the STOPPED state. Interface
• SCF ABORT CLIM or STOP CLIM
command
NonStop OS-initiated CLIM stop
resources still on the CLIM become
unavailable. Those resources that have
already failed over to other CLIMs
remain active unless the INTFALL option
is given.
Heartbeat time enforces sending a heartbeat signal from each CLIM to indicate it is running properly.
Failover is triggered only when all NonStop system processors have lost contact with the CLIM.
Link pulse time keeps failover from triggering because of a transient link pulse problem. It is a fixed
value less than two seconds.
All changes in the CLIM interface status are reported as EMS messages even if they do not result
in failover.
Each individual interface can be configured with its own failover interface, including the CLIM and
interface names. Failover interfaces are configured in pairs, so both interfaces of the pair specify
either the other as their failover interface or no failover. It is not necessary that all interfaces in a
CLIM have their failover interfaces on the same CLIM. It is also acceptable for a failover interface
to be a bonded interface or a physical interface regardless of the home-interface type.Table 1:
Interface Resource Failover Behavior (page 40) summarizes how CLIM-to-CLIM failover migrates
each type of interface resource from the failed interface to the failover interface. “Interface Resource
Migration” (page 45) describes migration of each of these resources in more detail. A two-way
periodic heartbeat timeout is used on both the NonStop host system and CLIM to detect failure of
a CLIM, the host, or the ServerNet connection between them. If a CLIM can no longer communicate
with any processor in the NonStop host system, it removes all the IP addresses from its data LAN
ports. If the NonStop host system is truly down, a down-state is conveyed to the external network.
If the NonStop host system is actually still up, it allows the host to migrate the IP addresses to
another CLIM. If no NonStop host system processor can communicate with a CLIM, the host initiates
a CLIM-to-CLIM failover of all the interfaces on the CLIM. The resources for each interface are
migrated to its configured failover interface. Each interface can have its own failover destination
and hence a CLIM failover could result in failover of interfaces to multiple CLIMs. Each interface
migrates separately from the others. Hence, a problem migrating one interface does not affect the
others.
Fail Back (Restore Behavior)
Fail back or restore is only supported by a manual method and initiated from the host only by
using a SWITCH CLIM command. Fail back of interfaces also follows the “check-break-make”
principle and the actual sequence is similar to a failover sequence. See “Failover in the CIP
Subsystem” (page 36) for an overview of the CIP failover sequence.
44 Overview