The task of this cutover is to ensure that certain servers stay connected to the Internet as much as possible, replace the core switch, and put the new machine on the shelf. In view of the above problems, the solution is analyzed
. The two PE routers in the current network architecture are used as ASBR equipment and RR. Equipment, using the cross-domain solution C2, the core equipment as the CE equipment and PE to establish a BGP neighbor relationship. The two cores in the LAN interact with each other in VRRP, and the server switch is connected to the VRRP-master via a single link. And the introduction of direct and static routers in bgp The 
above topology is to simplify the topology. In the above structure, PE1 and PE2 establish the neighbor relationship of BGP member V4, PE1 and core-1 establish the neighbor relationship of BGP instance, the test device is this time In order to ensure that the server is continuously connected to the network and temporarily connected to the network, a switch and the PE device establish an instance neighbor relationship, and introduce direct connections and static routes. And it is interconnected with the CORE-1 switch, as a server gateway VRRP slave device, and core-1 core-2 negotiated vrrp
as follows to configure the test switch:
interface Vlanif10
ip binding -instance YW
ip address 10.1.14.2 255.255.255.0
#
interface Vlanif20
ip binding instance YW
ip address 20.1.1.4 255.255.255.0
vrrp vrid 20 virtual-ip 20.1.1.1
vrrp vrid 20 priority 150
bgp 200
router-id 10.1.14.2
#
ipv4-family unicast
undo synchronization
#
ipv4-family instance YW
import-route direct
peer 10.1.14.1 as-number 100
The above figure shows the PE-1 bgp neighbor situation and the private route acceptance situation. It can be seen from the above figure that the current router accepts routing and selects test, for which the MED of test is smaller than CORE-1
Now PE-1 selects the test device, and the test device is also added to the VRRP group as a slave device. Now we test whether the server can communicate normally (from the pe-1 PING server address 20.1.1.5) and find that PE chooses the path as PE1- TEST-CORE-1-server switch.
In the next step, we will switch to minimize the time reduction. The server cannot communicate temporarily due to the switch.
According to the current environment, the test device, as a slave device, can normally learn the MAC of the server and update the route to the router. According to the BGP update principle, only the optimal route is updated, and the route learned from core-1 will not be updated to neighbors. .
So we switch can be divided into two methods, the first is to directly turn test into the master device of vrrp, so PE1 will select PE1-TEST-server switch when running traffic, so that almost no interruption switching can be achieved. Because VRRP protocol 1s switch between master and backup, and send free arp after master switch to inform downstream switch
It can be seen that VRRP switching is very fast.
The second method is a more robust method is that we connect the test to the router and core-1, the server switch respectively. The premise is that we need to prepare the stp protocol in advance, so that we only need to gradually modify the MED and the test VRRP. Because we are a H3C device, stp is not turned on by default to avoid unnecessary troubles, so choose the first one, and I won’t repeat it here.