四、多节点部署
Lab 示例
多节点部署的实验演示
现在让我们构建一个 Lab，综合运用前面学习过的 IGP 与 iBGP 协议。拓扑结构如下图所示：
细心的你可能已经注意到，此处我们采用的是回环地址（Loopback IP）。这是因为在部署 iBGP 时，有一个被广泛认可的最佳实践：在存在 IGP 的网络环境中，建议使用回环地址作为 iBGP 的建立地址。这样一来，即便节点间的物理链路发生变化，IGP 也能及时收敛并更新路径，实现协议层次的合理分离。
当然，若两个节点之间是点对点直连（PtP），也可以直接使用接口地址建立 iBGP 会话，这在简单拓扑中也非常常见。
本实验采用 RR（Route Reflector）配置方式，指定 Node A 作为 RR 节点。functions.conf 等基础配置沿用我们在前一章节中的内容，此处不再赘述。
配置
IGP
protocol ospf v3 ospfv3 {
ipv6;
area 0 {
interface "e1" {
cost 10;
type ptp;
};
interface "e2" {
cost 10;
type ptp;
};
interface "dummy0" {
stub;
};
};
}protocol ospf v3 ospfv3 {
ipv6;
area 0 {
interface "e0" {
cost 10;
type ptp;
};
interface "dummy0" {
stub;
};
};
}protocol ospf v3 ospfv3 {
ipv6;
area 0 {
interface "e0" {
cost 10;
type ptp;
};
interface "dummy0" {
stub;
};
};
}eBGP
protocol bgp upstream {
local fe80::a as 114514;
neighbor fe80::aa%e0 as 65001;
direct;
ipv6 {
import where import_filter_upstream();
export where export_filter_upstream();
};
graceful restart;
};protocol bgp upstream {
local fe80::c as 114514;
neighbor fe80::cc%e0 as 65001;
direct;
ipv6 {
import where import_filter_upstream();
export where export_filter_upstream();
};
graceful restart;
};iBGP
template bgp ibgp {
local 2001:db8::a as 114514;
multihop 2;
ipv6 {
import all;
export all;
};
rr client;
graceful restart;
}
protocol bgp node_b from ibgp {
neighbor 2001:db8::b as 114514;
}
protocol bgp node_c from ibgp {
neighbor 2001:db8::c as 114514;
}protocol bgp node_a {
local 2001:db8::b as 114514;
neighbor 2001:db8::a as 114514;
multihop 2;
ipv6 {
import all;
export all;
};
graceful restart;
}protocol bgp node_a {
local 2001:db8::c as 114514;
neighbor 2001:db8::a as 114514;
multihop 2;
ipv6 {
import all;
export all;
};
graceful restart;
}全部配置
include "functions.conf";
log syslog all;
router id 1.1.1.1;
protocol device {
};
protocol kernel {
ipv6 {
export all;
};
};
protocol ospf v3 ospfv3 {
ipv6;
area 0 {
interface "e1" {
cost 10;
type ptp;
};
interface "e2" {
cost 10;
type ptp;
};
interface "dummy0" {
stub;
};
};
}
protocol bgp upstream {
local fe80::a as 114514;
neighbor fe80::aa%e0 as 65001;
direct;
ipv6 {
import where import_filter_upstream();
export where export_filter_upstream();
};
graceful restart;
};
template bgp ibgp {
local 2001:db8::a as 114514;
multihop 2;
ipv6 {
import all;
export all;
};
rr client;
graceful restart;
}
protocol bgp node_b from ibgp {
neighbor 2001:db8::b as 114514;
}
protocol bgp node_c from ibgp {
neighbor 2001:db8::c as 114514;
}log syslog all;
router id 1.1.1.2;
protocol device {
};
protocol kernel {
ipv6 {
export all;
};
};
protocol ospf v3 ospfv3 {
ipv6;
area 0 {
interface "e0" {
cost 10;
type ptp;
};
interface "dummy0" {
stub;
};
};
}
protocol static {
ipv6;
route 2001:db8:100::/48 reject {bgp_large_community.add((114514,1,3));};
}
protocol bgp node_a {
local 2001:db8::b as 114514;
neighbor 2001:db8::a as 114514;
multihop 2;
ipv6 {
import all;
export all;
};
graceful restart;
}log syslog all;
router id 1.1.1.3;
protocol device {
};
protocol kernel {
ipv6 {
export all;
};
};
protocol ospf v3 ospfv3 {
ipv6;
area 0 {
interface "e0" {
cost 10;
type ptp;
};
interface "dummy0" {
stub;
};
};
}
protocol bgp upstream {
local fe80::c as 114514;
neighbor fe80::cc%e0 as 65002;
direct;
ipv6 {
import where import_filter_upstream();
export where export_filter_upstream();
};
graceful restart;
};
protocol bgp node_a {
local 2001:db8::c as 114514;
neighbor 2001:db8::a as 114514;
multihop 2;
ipv6 {
import all;
export all;
};
graceful restart;
}验证
root@debian:~# birdc s r
BIRD 2.17.1 ready.
Table master6:
2001:db8:aa::/48 unreachable [static1 04:33:32.394] * (200)
2001:db8:100::/48 unicast [bgp1 04:44:21.386] * (100) [AS114514i]
via fe80::a on ens3
root@debian:~#root@debian:~# birdc s r
BIRD 2.17.1 ready.
Table master6:
2001:db8::b/128 unicast [ospfv3 04:44:17.618] * I (150/10) [1.1.1.2]
via fe80::525b:9dff:fe00:100 on ens4
unicast [node_b 04:44:21.087 from 2001:db8::b] (100/10) [i]
via fe80::525b:9dff:fe00:100 on ens4
unicast [node_c 04:48:37.945 from 2001:db8::c] (100/10) [i]
via fe80::52dd:ddff:fe00:200 on ens5
2001:db8:aa::/48 unicast [upstream 04:38:54.271] * (100) [AS65001i]
via fe80::aa on ens3
2001:db8:cc::/48 unicast [node_c 04:50:54.638 from 2001:db8::c] * (100/10) [AS65002i]
via fe80::52dd:ddff:fe00:200 on ens5
2001:db8::c/128 unicast [ospfv3 04:48:36.618] * I (150/10) [1.1.1.3]
via fe80::52dd:ddff:fe00:200 on ens5
unicast [node_b 04:48:37.031 from 2001:db8::b] (100/10) [i]
via fe80::525b:9dff:fe00:100 on ens4
unicast [node_c 04:48:37.945 from 2001:db8::c] (100/10) [i]
via fe80::52dd:ddff:fe00:200 on ens5
2001:db8::a/128 unicast [ospfv3 04:41:48.619] * I (150/0) [1.1.1.1]
dev dummy0
unicast [node_b 04:44:21.087 from 2001:db8::b] (100/10) [i]
via fe80::525b:9dff:fe00:100 on ens4
unicast [node_c 04:48:37.945 from 2001:db8::c] (100/10) [i]
via fe80::52dd:ddff:fe00:200 on ens5
2001:db8:100::/48 unicast [node_b 04:44:21.087 from 2001:db8::b] * (100/10) [i]
via fe80::525b:9dff:fe00:100 on ens4
root@debian:~#root@debian:~# birdc s r
BIRD 2.17.1 ready.
Table master6:
2001:db8::b/128 unicast [ospfv3 04:43:10.297] * I (150/0) [1.1.1.2]
dev dummy0
unicast [node_a 04:44:21.515 from 2001:db8::a] (100/10) [i]
via fe80::52fe:b6ff:fe00:301 on ens3
2001:db8:aa::/48 unicast [node_a 04:44:21.515 from 2001:db8::a] * (100/10) [AS65001i]
via fe80::52fe:b6ff:fe00:301 on ens3
2001:db8:cc::/48 unicast [node_a 04:50:55.068 from 2001:db8::a] * (100/20) [AS65002i]
via fe80::52fe:b6ff:fe00:301 on ens3
2001:db8::c/128 unicast [ospfv3 04:48:37.459] * I (150/20) [1.1.1.3]
via fe80::52fe:b6ff:fe00:301 on ens3
unicast [node_a 04:48:37.048 from 2001:db8::a] (100/10) [i]
via fe80::52fe:b6ff:fe00:301 on ens3
2001:db8::a/128 unicast [ospfv3 04:44:17.459] * I (150/10) [1.1.1.1]
via fe80::52fe:b6ff:fe00:301 on ens3
unicast [node_a 04:44:21.515 from 2001:db8::a] (100/10) [i]
via fe80::52fe:b6ff:fe00:301 on ens3
2001:db8:100::/48 unreachable [static1 04:43:10.196] * (200)
root@debian:~#root@debian:~# birdc s r
BIRD 2.17.1 ready.
Table master6:
2001:db8::b/128 unicast [ospfv3 04:48:38.195] * I (150/20) [1.1.1.2]
via fe80::52fe:b6ff:fe00:302 on ens3
unicast [node_a 04:48:38.547 from 2001:db8::a] (100/10) [i]
via fe80::52fe:b6ff:fe00:302 on ens3
2001:db8:aa::/48 unicast [node_a 04:48:38.547 from 2001:db8::a] * (100/10) [AS65001i]
via fe80::52fe:b6ff:fe00:302 on ens3
2001:db8:cc::/48 unicast [upstream 04:50:55.240] * (100) [AS65002i]
via fe80::cc on ens4
2001:db8::c/128 unicast [ospfv3 04:48:23.195] * I (150/0) [1.1.1.3]
dev dummy0
unicast [node_a 04:48:38.547 from 2001:db8::a] (100/10) [i]
via fe80::52fe:b6ff:fe00:302 on ens3
2001:db8::a/128 unicast [ospfv3 04:48:38.195] * I (150/10) [1.1.1.1]
via fe80::52fe:b6ff:fe00:302 on ens3
unicast [node_a 04:48:38.547 from 2001:db8::a] (100/10) [i]
via fe80::52fe:b6ff:fe00:302 on ens3
2001:db8:100::/48 unicast [node_a 04:48:38.547 from 2001:db8::a] * (100/20) [i]
via fe80::52fe:b6ff:fe00:302 on ens3
root@debian:~#root@debian:~# birdc s r
BIRD 2.17.1 ready.
Table master6:
2001:db8:cc::/48 unreachable [static1 04:49:59.628] * (200)
2001:db8:100::/48 unicast [bgp1 04:50:55.368] * (100) [AS114514i]
via fe80::c on ens3
root@debian:~#可以看到， Node B 上广播的路由已经成功地被发送到了 Upstream A 和 C，并且能够按照最优路径前往 Upstream A 和 C 的路由。
后续
你可能会好奇：如果网络规模更大，两个有 BGP 会话的节点之间并非直连，而是需要经过多个中间节点进行转发，这种情况下该如何处理？
在当前场景下，可以通过设置 RR，使中间节点同样获得 BGP 路由表，从而具备中转能力。但如果中间节点的内存资源较为紧张，无法完整承载 BGP 全表，又该怎么办？
这时候，我们就需要引入一种名为 MPLS（多协议标签交换） 的技术。MPLS 能够有效解决这一问题，不过它属于进阶内容，因此不会在本系列入门教程中展开介绍（TODO）。如果你对此感兴趣，可以自行查阅资料或者等待我将高级教程“咕咕咕”出来 。