124 Chapter 5: Intradomain IP Control Plane: Restarting IS-IS Gracefully
IETF IS-IS Restart Operation
This section describes the operation of the IETF IS-IS restart mechanism. For clarity, the
starting and restarting router cases are described separately. Figure 5-6 shows the network
diagram for this example. Both routers are assumed to be IETF IS-IS restart- and NSF-capable
(able to preserve IP forwarding state across IS-IS restart). R1 and R2 are interconnected via
point-to-point links, and the point-to-point adjacencies are established using the three-way
handshake procedure. The sequence of steps described in the next section might not necessarily
be followed by a specific implementation. These steps are just meant to provide a functional-
level description of the IETF IS-IS restart mechanism.
Starting Router Operation
Figure 5-6 shows the network diagram for the starting router case, and the corresponding
sequence of steps follows.
Figure 5-6 Example of the IETF IS-IS Restart Operation for the Starting Router Case
1.
Assume R1 is running normally. (That is, LSP databases have been fully synchronized.)
2. Suppose R2 starts.
R1
R2
R1 runs normally. R2 starts.
Sends an IIH with the SA bit set and the RR bit clear.
Normal operation resumes.
Sends IIH packet with the RA bit set (RR and SA bits clear), schedules
transmission of CSNPs, and continues to suppress advertisement
of adjacency.
R1 synchronizes the LSP database, upon completion cancels T2 timers,
clears the OL bits, updates RIB/FIB, and sends IIH with the RR, RA,
SA bits clear.
1
2
3
4
5
6
7
8
9
R1 establishes a new adjacency or re-initializes the existing adjacency
and suppresses advertisement of the adjacency.
Sends IIH with the SA and RR bit set. Before CSNPs,
sends zeroth LSP with the OL bits set.
R1 begins to advertise the adjacency and reconsiders it in
SPF calculations.
IETF IS-IS Restart Operation 125
3.
On starting, R2 activates the T2 timer for each level database and establishes adjacencies
by transmitting IIH packets containing a restart TLV with the SA bit set and the RR bit
clear.
4. On receipt of the IIH packet with the SA bit set and the RR bit clear, R1 follows the normal
procedure for an adjacency establishment. That is, if there is no existing adjacency on the
concerned interface, R1 establishes a new adjacency. Otherwise, it re-initializes the
existing adjacency. Unlike the normal procedure, however, R1 suppresses advertisement
of the adjacency in its LSPs and, to prevent temporary black holes, it excludes that
adjacency from the SPF calculations.
5. When an adjacency transitions to the up state, R2 starts the T1 timer and transmits an IIH
packet with the SA bit set and the RR bit clear on that interface. When the T1 timer expires
for the first time, it is restarted and an IIH packet is retransmitted with both the SA and
RR bits set. As a result of having sent an IIH packet with the RR bit set, R2 expects to
receive both an acknowledgment and a set of CSNPs from the neighbor over the
concerned interface. Because R2 knows that its database is incomplete, immediately after
establishing the adjacency but before the CSNPs exchange, it transmits the zeroth LSP
with the OL bits set. This ensures other routers (including R1) in the domain remove R2
from their forwarding paths.
6. Upon receipt of the IIH packet with the RR bit set, R1 maintains the established adjacency
state, sends an IIH packet with the RA bit set (RR and SA bits clear), sets SRMflags, sets
the restart mode flag (if previously clear), updates the hold time, and schedules
transmission of a complete set of CSNPs on that interface. In addition, because the SA bit
is set in the received IIH packet, R2 continues to suppress the adjacency advertisement and
excludes it from the SPF calculations.
7. Upon receiving both an acknowledgment and a set of CSNPs, R2 cancels the associated
T1 timer. Now R1 and R2 exchange LSPs and synchronize their databases. When the
database synchronization is complete at both levels, R2 cancels all T2 timers and
simultaneously clears both OL bits. From this point on, R2 transmits IIH packets with the
RR, RA, and SA bits clear. Irrespective of T2 timers, as R2 learns new routes it runs the
SPF algorithm and updates the RIB and the FIB.
8. Upon receiving the RR bit cleared, R1 clears the restart mode flag. As in the received IIH
packet, the SA bit is clear (indicating the end of adjacency suppression), and R1 begins to
advertise the concerned adjacency in its own LSPs and reconsiders it in the SPF
calculations.
9. Normal operation resumes.

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

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.