для Cisco Cisco IOS Software Release 12.0(29)S

Скачать
Страница из 38
      MPLS Traffic Engineering—RSVP Graceful Restart
Information About MPLS TE—RSVP Graceful Restart
3
Cisco IOS Release 12.0(29)S
Figure 1
How Graceful Restart Works
The Restart_Cap object has two values—the restart time, which is the sender's time to restart the 
RSVP_TE component and exchange hello messages after a failure; and the recovery time, which is the 
desired time that the sender wants the receiver to synchronize the RSVP/MPLS databases. 
In 
, graceful restart is enabled on Router 1, Router 2, Router 3, and Router 4. For simplicity, 
assume that all routers are restart capable. A TE label-switched path (LSP) is signaled from Router 1 to 
Router 4.
Router 2 and Router 3 exchange periodic graceful restart hello messages every 10000 milliseconds (10 
seconds), and so do Router 2 and Router 1 and Router 3 and Router 4. Assume that Router 2 advertises 
its restart time = 60000 milliseconds (60 seconds) and its recovery time = 60000 milliseconds (60 
seconds) as shown below: 
23:33:36: Outgoing Hello:
23:33:36:   version:1 flags:0000 cksum:883C ttl:255 reserved:0 length:32
23:33:36:  HELLO                type HELLO REQUEST length 12:
23:33:36:   Src_Instance: 0x6EDA8BD7, Dst_Instance: 0x00000000
23:33:36:  RESTART_CAP          type 1 length 12:
23:33:36:   Restart_Time: 0x0000EA60, Recovery_Time: 0x0000EA60
Router 3 records this into its database. Also, both neighbors maintain the neighbor status as UP. 
However, Router 3’s control plane fails at some point (for example, a Primary Route Processor failure). 
As a result, RSVP and TE lose their signaling information/states although data packets are kept 
forwarded by the line cards. 
When four ACK messages are missed from Router 2 (40 seconds), Router 3 declares communication 
with Router 2 lost indicated by LOST and starts the restart time to wait for the duration advertised in 
Router 2’s restart time previously and recorded (60 seconds). Router 1 and Router 2 suppress all RSVP 
messages to Router 3 except hellos. Router 3 keeps sending the RSVP Path and Resv refresh messages 
to Router 4 and Router 5 so that they do not expire the state for the LSP; however, Router 3 suppresses 
these messages for Router 2.
Note
A node is restarting if it misses four ACKs or its hello src_instance (last source instance sent to its 
neighbor) changes so that its restart time = 0.
Before the restart time expires, Router 2 restarts and loads its configuration and graceful restart makes 
it send the hello messages with a new source instance to all the datalinks attached. However, since Router 
2 has lost the neighbor states, it does not know what destination instance it should use in those messages; 
therefore, all destination instances are set to 0. 
When Router 3 sees the hello from Router 2, Router 3 stops the restart time for Router 2 and sends an 
ACK message back. When Router 3 sees a new source instance value in Router 2’s hello message, Router 
3 knows that Router 2 had a control plane failure. Router 2 gets Router 3’s source instance value and 
uses it as the destination instance going forward.
117933
Router 1
SSO Help Neighbor
Router 2
Hello Restart_Cap
Router 3
Tunnel
Full SSO Help
Router 5
Router 4
Hello Restart_Cap