当前位置：首页 > news >正文

M-LAG与E-trunk

news 2026/2/8 20:13:00

M-LAG和E-trunk都是用来实现跨设备链路聚合，解决单点故障的，其大部分特性相同，工作模式M-LAG更胜一筹，支持双活,而且其原理感觉像是vrrp+mstp的升级版，是往增加网络可靠性去发展的;而E-trunk是基于LACP扩展实现，感觉主要就是为了实现跨设备链路聚合而创造的链路聚合拓展协议，然后再进行扩展得到的，个人感觉M-LAG更灵活便捷

M-LAG

M-LAG跨设备链路聚合是一种靠可用技术，其相较于堆叠有更高的灵活性

DRCP

Distributed Relay Control Protocol，分布式聚合控制协议

该协议跑在peer-link上，其目的是为完成M-LAG组成员间的信息交互，当本端接口超时时仍未收到对端的DRCP报文，则认为peer-link为down状态

DRCP超时时间可选短超时（3s），此时发送周期1s，也可选长超时（90s），此时发送周期30s

M-LAG接口

即M-LAG的下行接口，一般有单归接入和双归接入

单归接入，即一台设备接入单台M-LAG系统设备，也叫单挂，单挂的设备在M-LAG系统中，其Mac地址表，ARP表等都会进行备份，依次给其留下备份路径提高可靠性

双归接入，一台设备接入两台设备，上联的这两台设备可实现负载分担，故障可快速回切

keepalive

用于检测keepalive-link的活动状态，周期性发送keepalive报文，收到对端认为keepalive-link为up，反之为down

peer-link

此链路为UP，表明M-LAG正常工作，如果此时keepalive-link为down，系统正常工作，但会打印日志通知管理员检测

此链路为Down，则使用keepalive-link收到的报文进行选举主设备，保障系统能正常转发数据

当peer-link为Down时，认为对端设备挂了，启动keepalive timeout计时器，计时器超时后有以下情况：

设备为主设备，有M-LAG接口为UP，设备为主，否则为None

设备为从设备，升主，之后有M-LAG接口为UP，设备保持为主，否则切为None

设备为None，设备不能收发keepalive报文，keepalive-link处于down状态

角色计算

说了这么多，其实就这一张图，和堆叠类似

比较M-LAG接口状态，有可工作的优先
先前状态
不在MAD Down状态的优先
设备健康值，越小越优
比较优先级，越大越优先，
比较MAC，越小越优先

防环机制

设备A为非M-LAG接口，怎么走？

假如主设备为E，A发包给D，D发给E，丢包？？？

本地转发优先

接收流量的设备存在表项，只在此设备上发送不经peer-link转发

下图，设备B链路负载分担，怎么走？

好根据上述原则，假设D有，DE是负载分担，那我有一半流量要绕peer-link，这就是缺陷

环路？？？

单向转发避免

从peer-link收到的消息不向M-LAG接口发送

MAD检测

为防止设备peer-link链路故障后设备间重新选举将流量误导，配置MAD Down之后，当故障发生时会关闭从设备除了管理员指定的所有接口，设备进入MAD Down状态

此时如果keepalive-link故障，则从设备晋升为主，网络中存在两台主设备，可能引发二次故障，可以开启MAD Down保持状态进行干预

一致性检测

为确保两端设备匹配，不影响相关报文转发对相关数据进行一致性检验，目前一般两种：

Type 1类型配置：影响M-LAG系统转发的配置，如果Type 1类型配置不匹配，则将从设备上M-LAG接口置为down状态，比如vlan的配置

Type 2类型配置：仅影响业务模块的配置，如果Type 2类型配置不匹配，从设备上M-LAG接口依然为up状态，不影响M-LAG系统正常工作，由Type 2类型配置对应的业务模块决定是否关闭该业务功能，其他业务模块不受影响，比如web认证，端口安全模式

为了避免设备 M-LAG 接口震荡，设备会在延迟恢复定时器（缺省为30s）一半时间之后进行配置一致性检查

双活网关

两端配置相同vlan以及ip，用双归接入，其实这个拓扑和堆叠也就差个原理罢了，简化不就是链路聚合+堆叠吗？这么接好处心中有数，可实现负载分担快速切换

总结

经过DRCP协商交换配置和资源，构建M-LAG系统，进行主从协商检测一致性，发送keepalive报文检测邻居状态是否正常，正常后进行数据同步与信息交互，M-LAG开始运行，peer-link决定M-LAG系统的工作状态，keepalive用于检测对端状态是否正常，结合peer-link状态进行判断，设备间通过信息交互实现高可靠性

实验

拓扑

配置

1.m-lag的虚拟mac地址优先级


[sw1]m-lag system-number 1
Changing the system number might flap the peer link and cause M-LAG system setup failure. Continue? [Y/N]:y
[sw1]%Jan 22 22:15:47:060 2025 sw1 M-LAG/6/MLAG_SYSEVENT_NUMBER_CHANGE: System number changed from default to 1.
[sw1]m-lag system-mac 1-1-1
Changing the system MAC address might flap the peer link and cause M-LAG system setup failure. Continue? [Y/N]:y
[sw1]%Jan 22 22:16:16:894 2025 sw1 M-LAG/6/MLAG_SYSEVENT_MAC_CHANGE: System MAC address changed from N/A to 0001-0001-0001.
[sw1]m-lag system-priority 120
Changing the system priority might flap the peer link and cause M-LAG system setup failure. Continue? [Y/N]:y
[sw1]%Jan 22 22:18:11:451 2025 sw1 M-LAG/6/MLAG_SYSEVENT_PRIORITY_CHANGE: System priority changed from 32768 to 120.

2.keepalive链路

[sw1-Ten-GigabitEthernet1/0/50]port link-mode route
[sw1-Ten-GigabitEthernet1/0/50]ip add 10.0.0.1 24
[sw1-Ten-GigabitEthernet1/0/50]qui
[sw1]m-lag keepalive ip destination 10.0.0.2 source 10.0.0.1
[sw1]m-lag mad exclude int Ten-GigabitEthernet 1/0/5

3.peer-link链路

[sw1]int Bridge-Aggregation 1
[sw1-Bridge-Aggregation1]link-aggregation mode dynam
[sw1]int Ten-GigabitEthernet 1/0/51
[sw1-Ten-GigabitEthernet1/0/51]port link-aggregation group 1
[sw1-Ten-GigabitEthernet1/0/51]%Jan 22 22:27:58:073 2025 sw1 LAGG/6/LAGG_LACP_RECEIVE_TIMEOUT: LACPDU reception timed out on member port XGE1/0/51 in aggregation group BAGG1.
%Jan 22 22:27:58:082 2025 sw1 IFNET/5/LINK_UPDOWN: Line protocol state on the interface Ten-GigabitEthernet1/0/51 changed to down.
int t%Jan 22 22:28:02:030 2025 sw1 LAGG/6/LAGG_ACTIVE: Member port XGE1/0/51 of aggregation group BAGG1 changed to the active state.
%Jan 22 22:28:02:036 2025 sw1 STP/6/STP_NOTIFIED_TC: Instance 0's port Ten-GigabitEthernet1/0/52 was notified a topology change.
%Jan 22 22:28:02:040 2025 sw1 STP/6/STP_DETECTED_TC: Instance 0's port Bridge-Aggregation1 detected a topology change.
%Jan 22 22:28:02:041 2025 sw1 IFNET/5/LINK_UPDOWN: Line protocol state on the interface Ten-GigabitEthernet1/0/51 changed to up.
%Jan 22 22:28:02:042 2025 sw1 IFNET/3/PHY_UPDOWN: Physical state on the interface Bridge-Aggregation1 changed to up.
%Jan 22 22:28:02:042 2025 sw1 IFNET/5/LINK_UPDOWN: Line protocol state on the interface Bridge-Aggregation1 changed to up.
en
[sw1-Ten-GigabitEthernet1/0/51]int ten1/0/52
[sw1-Ten-GigabitEthernet1/0/52]port link-aggregation g 1
[sw1-Ten-GigabitEthernet1/0/52]int bri 1
[sw1-Bridge-Aggregation1]port m-lag peer-link 1
[sw1]m-lag consistency-check disable

4.m-lag接口


[sw1]int Bridge-Aggregation 2
[sw1-Bridge-Aggregation2]link-aggregation mode dynamic
[sw1-Bridge-Aggregation2]int g1/0/1
[sw1-GigabitEthernet1/0/1]port link-aggregation g 2
[sw1]int Bridge-Aggregation 2
[sw1-Bridge-Aggregation2]port m-lag group 2

5.对端配置

注意：记得关闭一致性检测（模拟器环境真机不推荐）,关闭静态mac源检测（在pee，记得关闭，不然下面接口起不来

[sw2]m-lag system-mac 1-1-1
Changing the system MAC address might flap the peer link and cause M-LAG system setup failure. Continue? [Y/N]:y
[sw2]%Jan 22 22:37:01:457 2025 sw2 M-LAG/6/MLAG_SYSEVENT_MAC_CHANGE: System MAC address changed from N/A to 0001-0001-0001.[sw2]m-lag system-priority 120
Changing the system priority might flap the peer link and cause M-LAG system setup failure. Continue? [Y/N]:y
[sw2]%Jan 22 22:37:15:524 2025 sw2 M-LAG/6/MLAG_SYSEVENT_PRIORITY_CHANGE: System priority changed from 32768 to 80.[sw2]m-lag system-number 2
Changing the system number might flap the peer link and cause M-LAG system setup failure. Continue? [Y/N]:y
[sw2]%Jan 22 22:37:27:386 2025 sw2 M-LAG/6/MLAG_SYSEVENT_NUMBER_CHANGE: System number changed from default to 2.[sw2]int Ten-GigabitEthernet 1/0/50
[sw2-Ten-GigabitEthernet1/0/50]port link-mode route
[sw2-Ten-GigabitEthernet1/0/50]ip address 10.0.0.2 24
[sw2]m-lag keepalive ip destination 10.0.0.1 source 10.0.0.2
[sw2]m-lag mad exclude int Ten-GigabitEthernet 1/0/50[sw2]int Bridge-Aggregation 1
[sw2-Bridge-Aggregation1]link-aggregation mode dynamic[sw2-Bridge-Aggregation1]int ten1/0/51
[sw2-Ten-GigabitEthernet1/0/51]port link-aggregation g 1
%Jan 22 22:43:58:431 2025 sw2 LAGG/6/LAGG_LACP_RECEIVE_TIMEOUT: LACPDU reception timed out on member port XGE1/0/51 in aggregation group BAGG1.
[sw2-Ten-GigabitEthernet1/0/51]int t1/0/52
[sw2-Ten-GigabitEthernet1/0/51]port link-aggregation g 1[sw2]int Ten-GigabitEthernet1/0/52
[sw2-Ten-GigabitEthernet1/0/52]port link-aggregation g 1
[sw2-Ten-GigabitEthernet1/0/52]int bri 1
[sw2-Bridge-Aggregation1]port m-lag peer-link 1[sw2]int bri 2
[sw2-Bridge-Aggregation2]p
[sw2-Bridge-Aggregation2]link-ag
[sw2-Bridge-Aggregation2]link-aggregation mode dy
[sw2-Bridge-Aggregation2]link-aggregation mode dynamic
[sw2-Bridge-Aggregation2]int g1/0/1
[sw2-GigabitEthernet1/0/1]port link-ag g 2[sw2-GigabitEthernet1/0/1]int bri 2
[sw2-Bridge-Aggregation2]port m-lag group 2
[sw2]m-lag consistency-check disable

下接设备做链路聚合


[sw3]int bri 2
[sw3-Bridge-Aggregation2]link mode dy
[sw3-Bridge-Aggregation2]link mode dy
[sw3-Bridge-Aggregation2]int ran g1/0/1 g1/0/2
[sw3-if-range]port link g 2

看下keepalive，没有问题

看下sw3的链路聚合，两条都起来了，也没有问题

E-trunk

LACP

链路聚合原理，不过多解释，看前面的笔记

系统ID：越小越优先，缺省为使用e-trunk的接口mac

E-Trunk的优先级：用于在聚合组中决策两台设备的主备状态，越小越优

E-Trunk的ID：唯一标识号

主备协商

CE分别与PE1和PE2直连，PE1和PE2之间运行E-Trunk

PE侧

在PE1和PE2设备上分别创建ID相同的E-Trunk和Eth-Trunk，并将Eth-Trunk加入到E-Trunk

CE侧

在CE设备上配置LACP模式的Eth-Trunk，此Eth-Trunk分别与PE1和PE2设备相连

对CE设备而言，E-Trunk不可见

确定E-Trunk的主备状态

PE1与PE2设备之间通过E-Trunk报文进行主备协商，确定E-Trunk的主备状态，正常情况下两台PE的协商结果是一个为主用一个为备用。

PE设备上E-Trunk主备状态是根据报文中所携带的E-Trunk优先级和E-Trunk系统ID确定的，优先级的数值越小，优先级越高，优先级高的为主用。如果E-Trunk优先级相同，那么E-Trunk系统ID小的为主用
说白了就是欺骗设备告诉其绑定的是同一台设备

实验

配置

pe端，创建e-trunk组将eth-trunk加入e-trunk

[sw2]int eth 1
[sw2-Eth-Trunk1]mode lacp	
[sw2-Eth-Trunk1]mode lacp-static 
[sw2-Eth-Trunk1]int g0/0/2
[sw2-GigabitEthernet0/0/2]eth 1
Info: This operation may take a few seconds. Please wait for a moment...done.
[sw2-GigabitEthernet0/0/2]
Jan 23 2025 10:45:41-08:00 sw2 %%01IFNET/4/IF_STATE(l)[0]:Interface Eth-Trunk1 h
as turned into UP state.
Jan 23 2025 10:45:42-08:00 sw2 DS/4/DATASYNC_CFGCHANGE:OID 1.3.6.1.4.1.2011.5.25
.191.3.1 configurations have been changed. The current change number is 6, the c
hange loop count is 0, and the maximum number of records is 4095.
[sw2-GigabitEthernet0/0/2]e-tr	
[sw2-GigabitEthernet0/0/2]lacp e-trunk system-id 1-1-1
[sw2]lacp e-
Jan 23 2025 10:46:22-08:00 sw2 DS/4/DATASYNC_CFGCHANGE:OID 1.3.6.1.4.1.2011.5.25
.191.3.1 configurations have been changed. The current change number is 7, the c
hange loop count is 0, and the maximum number of records is 4095.trunk p	
[sw2]lacp e-trunk priority 12800
[sw2]e-trunk 1
[sw2]int eth 1
[sw2-Eth-Trunk1]e-t 1

[sw3]eth 1^
Error: Unrecognized command found at '^' position.
[sw3]int eth 1
[sw3-Eth-Trunk1]mode lacp	
[sw3-Eth-Trunk1]mode lacp-static 
[sw3-Eth-Trunk1]int g0/0/2
[sw3-GigabitEthernet0/0/2]
[sw3-GigabitEthernet0/0/2]
Jan 23 2025 10:49:13-08:00 sw3 DS/4/DATASYNC_CFGCHANGE:OID 1.3.6.1.4.1.2011.5.25
.191.3.1 configurations have been changed. The current change number is 5, the c
hange loop count is 0, and the maximum number of records is 4095.
[sw3-GigabitEthernet0/0/2]port eth	
[sw3-GigabitEthernet0/0/2]eth 1
Info: This operation may take a few seconds. Please wait for a moment...done.
[sw3-GigabitEthernet0/0/2]e
Jan 23 2025 10:49:43-08:00 sw3 DS/4/DATASYNC_CFGCHANGE:OID 1.3.6.1.4.1.2011.5.25
.191.3.1 configurations have been changed. The current change number is 6, the c
hange loop count is 0, and the maximum number of records is 4095.-trunk 	
[sw3-GigabitEthernet0/0/2]lacp e-trunk system-id 1-1-1
[sw3]lacp 
Jan 23 2025 10:50:13-08:00 sw3 DS/4/DATASYNC_CFGCHANGE:OID 1.3.6.1.4.1.2011.5.25
.191.3.1 configurations have been changed. The current change number is 7, the c
hange loop count is 0, and the maximum number of records is 4095.e-trunk	
[sw3]e-trunk 1
[sw3]lacp e-trunk p	
[sw3]lacp e-trunk priority 12800
[sw3]int eth 1
[sw3-Eth-Trunk1]e-t 1

ce段做链路聚合,整完之后可以看到，已经实现跨设备链路聚合

为保持可靠性可以采用bfd联动，在23间在加线来实现，这里我在23间在做个链路聚合

[sw2-Eth-Trunk1]e-t 1
[sw2-Eth-Trunk1]
Jan 23 2025 10:47:50-08:00 sw2 %%01ETRUNK/4/MBR_PHYSTAT_CHG(l)[1]:The physical s
tatus of an E-Trunk member changed. (MemberType=Eth-Trunk, MemberId=1, State=Dow
n)
Jan 23 2025 10:47:50-08:00 sw2 %%01IFNET/4/IF_STATE(l)[2]:Interface Eth-Trunk1 h
as turned into DOWN state.
Jan 23 2025 10:47:52-08:00 sw2 DS/4/DATASYNC_CFGCHANGE:OID 1.3.6.1.4.1.2011.5.25
.191.3.1 configurations have been changed. The current change number is 10, the 
change loop count is 0, and the maximum number of records is 4095.
[sw2-Eth-Trunk1]dis th
#
interface Eth-Trunk1mode lacp-statice-trunk 1
#
return
[sw2-Eth-Trunk1]qu
[sw2]int eth 2
[sw2-Eth-Trunk2]mode lacp
[sw2-Eth-Trunk2]int g0/0
Jan 23 2025 10:54:42-08:00 sw2 DS/4/DATASYNC_CFGCHANGE:OID 1.3.6.1.4.1.2011.5.25
.191.3.1 configurations have been changed. The current change number is 11, the 
change loop count is 0, and the maximum number of records is 4095./1
[sw2-GigabitEthernet0/0/1]eth 2
Info: This operation may take a few seconds. Please wait for a moment...done.
[sw2-GigabitEthernet0/0/1]int g0/0/
Jan 23 2025 10:54:52-08:00 sw2 DS/4/DATASYNC_CFGCHANGE:OID 1.3.6.1.4.1.2011.5.25
.191.3.1 configurations have been changed. The current change number is 12, the 
change loop count is 0, and the maximum number of records is 4095.3
[sw2-GigabitEthernet0/0/3]eth 2
Info: This operation may take a few seconds. Please wait for a moment...
Jan 23 2025 10:55:01-08:00 sw2 %%01IFNET/4/IF_STATE(l)[3]:Interface Vlanif1 has 
turned into DOWN state.done.

bfd的配置记得放行vlan，配置vlanif地址

bfd
qu
bfd 1
bfd 1 bind peer-ip 10.0.0.3 source-ip 10.0.0.2 auto
int eth 2
p l t
p t a v a