128 Chapter 5: Intradomain IP Control Plane: Restarting IS-IS Gracefully
11.
After restart, R2 regenerates its own LSPs, which should be the same as those generated
before this restart. Therefore, reception of R2’s LSPs should not trigger SPF calculations
on R1 and other routers in the domain.
12. Normal operation resumes.
Cisco IS-IS Restart
The IETF IS-IS restart mechanism extends the IS-IS protocol and requires that both the
restarting router and its neighbors be able to support these capabilities. In contrast, the Cisco
IS-IS restart mechanism does not make any externally observable protocol changes. That is, the
IS-IS restart on a router is hidden from its neighbors. First, this mechanism requires
transmission of IIH packets quickly following the restart before an adjacency’s hold time
expires. Second, to conceal the restart event following the restart, this mechanism relies on the
capability of the restarting router to transmit IIH packets containing exactly the same
information as existed before the restart. Therefore, to mask the IS-IS restart event from other
routers, this approach requires preservation of some IS-IS adjacency-related state information
across the restart.
Although the IETF restart mechanism uses protocol extensions to affect changes in the restart
behavior, the Cisco restart takes a different approach to satisfy the three objectives stated
previously:
First, to prevent SPF runs throughout the routing domain caused by a change in an
adjacency state, none of the existing up adjacencies between the restarting router and its
neighbors should be initialized across the restart.
Second, to ensure LSP database synchronization, the restarting router needs to have some
way to make neighbors set their SRMflags for all LSPs.
Third, to prevent temporary black holes that might result from incorrect or incomplete
information, the restarting router should have a reliable way to determine when its
database synchronization is complete.
The restarting routers supporting the Cisco IS-IS restart meet the first objective by transmitting
IIH packets (which appear the same as they were before restart) before an adjacency expires. A
router running with the Cisco IS-IS restart meets the second objective through transmission of
a CSNP containing the special LSP. However, this mechanism does not meet the third objective
because it cannot invoke transmission of CSNPs on point-to-point interfaces. Note the IETF IS-
IS restart achieves this through protocol extensions (RR and RA bits). In the absence of a
reliable way, the Cisco IS-IS restart uses heuristics such as monitoring the arrival rate of LSPs.
For example, when the arrival rate of LSPs falls below a certain predefined threshold value, the
database synchronization is deemed complete.
The Cisco IS-IS restart mechanism preserves some IS-IS adjacency-related state information
across the IS-IS restart. This information is needed to transmit IIH packets after the restart to
make IIH packets appear exactly the same to neighbors as before the restart. After restart, the

Get Fault-Tolerant IP and MPLS Networks now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.