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

the network. To specify a route for the first hop on the network in CIP, use the climconfig route
commands.
Failover Differences (Fault Tolerance)
There are two forms of failover in CIP: CLIM-to-CLIM failover and interface-to-interface failover.
CLIM-to-CLIM failover allows an interface on a CLIM to fail over to an interface on another CLIM.
Interface-to-interface failover is provided by the bonding feature. Use the CLIMCMD Climconfig
commands to establish failover between interfaces on a CLIM and between interfaces on two
different CLIMs.
Bonded interfaces share interface resources among multiple physical interfaces. They can be
configured to be similar to NonStop TCP/IPv6 and Parallel Library TCP/IP failover with the SHAREDIP
option, except that the interfaces must be in the same CLIM and are not limited to just two interfaces.
The NonStop TCP/IPv6 and Parallel Library TCP/IP NONSHAREDIP option is most similar
CLIM-to-CLIM failover in that each interface has a different IP address until failover with some
important differences.
While most resources can be migrated during failover, some are lost if migration to a different
CLIM is required. Unlike NonStop TCP/IPv6 and Parallel Library TCP/IP, CIP does not migrate
socketsthat have TCP or SCTP connections to the new CLIM. During a CLIM-to-CLIM failure, such
sockets are closed and TCP or SCTP sessions lost. (For a description of interface resources, see
“Failover in the CIP Subsystem” (page 36).
Restoring resources to their home interface when it comes back up is not done automatically and
requires you to issue an SCF SWITCH CLIM command. However, visiting resources using an
interface that fails do attempt restoration to their home interface automatically. The restore procedure
is similar to a failover sequence in how it handles interface resource migration. See “CLIM Startup
Behavior” (page 45) for information about how the CLIM manages recovery.
For procedures for setting up bonded interfaces and CLIM-to-CLIM failover, see “Configuring
Bonded Interface Failover (IP CIP)” (page 94).
Differences That Affect Planning for CLIMs Instead of G4SAs
In CIP, there are differences from conventional TCP/IP and NonStop TCP/IPv6 in how the physical
interfaces are divided among transport-service providers. These differences affect planning decisions
for the number of CLIMs needed to support some existing G4SA-based network configurations.
Each logical network partition (LNP) in NonStop TCP/IPv6 or conventional TCP/IP process can
have its own, independent set of resources (Ethernet interfaces). One interface on a G4SA can be
part of one partition or process while another interface can be part of another partition or interface.
The four different interfaces of one G4SA can belong to four different partitions or processes.
In CIP, each Provider can include more than one CLIM but a CLIM cannot be divided among
different Providers. Hence, all Ethernet interfaces on a CLIM can only belong to one Provider.
If you had a network configuration with a G4SA split between four LNPs or TCP/IP processes, one
interface on each, you may need to change your configuration set-up or add more CLIMs to achieve
the same result.
IPv6 Differences
Automatic tunnels are not supported in CIP.
Fault Tolerant Sockets
Conventional TCP/IP is a NonStop process pair and therefore allows sockets to be transferred
from an application in one processor to its backup on another. NonStop TCP/IPv6 and CIP do not
support this socket feature.
Operational Differences Between Previous TCP/IP Subsystems and CIP 183