华为M-Lag最佳实践

Server leaf的M-LAG配置

配置M-LAG


[~ServerLeaf1_1] 

 stp mode rstp

stp v-stp enable   //配置V-STP方式的M-LAG

stp tc-protection   //使能设备对TC类型BPDU报文的保护功能

stp bpdu-protection   //使能设备的BPDU保护功能

arp ip-conflict-detect enable   //使能设备的IP地址冲突检测的功能,网络虚拟化场景不需要配置

# 


[~ServerLeaf1_2] 
 stp mode rstp
 stp v-stp enable   //配置V-STP方式的M-LAG
 stp tc-protection              
 stp bpdu-protection           
 arp ip-conflict-detect enable  
 # 
 

配置M-LAG的DFS组


[~ServerLeaf1_1] 
ip vpn-instance DAD   //配置DAD VPN
 ipv4-family
  route-distinguisher 21:14
#
interface Eth-Trunk20   //配置双主检测链路,DAD成员口在框式设备上必须跨板部署
 trunkport 10GE 1/0/47 to 1/0/48
 mode lacp-static 
 undo portswitch
 ip binding vpn-instance DAD
 ip address 10.254.124.2 255.255.255.0   //配置source ip地址
 m-lag unpaired-port reserved
#
dfs-group 1 
 priority 150    //配置DFS优先级高于对端,默认是100 
 source ip 10.254.124.2 vpn-instance DAD peer 10.254.124.3   
 dual-active detection enhanced enable
#
 [~ServerLeaf1_2] 
ip vpn-instance DAD
 ipv4-family
  route-distinguisher 22:14
#
interface Eth-Trunk20
 trunkport 10GE 1/0/47 to 1/0/48
 mode lacp-static 
 undo portswitch
 ip binding vpn-instance DAD
 ip address 10.254.124.3 255.255.255.0
 m-lag unpaired-port reserved
# 
dfs-group 1 
 source ip 10.254.124.3 vpn-instance DAD peer 10.254.124.2   
 dual-active detection enhanced enable
#
 

配置peer-link

[~ServerLeaf1_1] 
 interface Eth-Trunk0   //如果是框式设备,则Peer-link成员口必须跨板部署,保证链路可靠性。如果单板速率不一致可配置lacp mixed-rate link enable来允许不同速率端口混合捆绑
  trunkport 40GE 1/0/5 to 1/0/6 
  mode lacp-static 
  peer-link 1 
 #
[~ServerLeaf1_2] 
 interface Eth-Trunk0 
  trunkport 40GE 1/0/5 to 1/0/6 
  mode lacp-static 
  peer-link 1 
 #

配置业务服务器以负载分担方式接入。

[~ServerLeaf1_1] 
interface eth-trunk 10 
   port link-type trunk              
   undo port trunk allow-pass vlan 1  
   trunkport 10ge 1/0/1 
   dfs-group 1 m-lag 10 
   mode lacp-static  
   stp edged-port enable    //配置边缘端口
 #
 interface 10GE1/0/1   //服务器接入端口
  storm suppression unknown-unicast 5   //配置未知单播抑制,经验值为10GE端口的5%带宽,建议业务端口都部署
  storm suppression multicast packets 1000  //配置组播报文抑制,接入VXLAN的为组播业务时不可以配置,单播流量可配置,经验值为1000pps
  storm suppression broadcast packets 1000  //配置广播报文抑制,经验值为1000pps,建议业务端口都部署
 #
[~ServerLeaf1_2] 
 interface eth-trunk 10 
   port link-type trunk  
   undo port trunk allow-pass vlan 1  
   trunkport 10ge 1/0/1 
   dfs-group 1 m-lag 10 
   mode lacp-static  
   stp edged-port enable 
 #
 interface 10GE1/0/1
   storm suppression unknown-unicast 5
   storm suppression multicast packets 1000
   storm suppression broadcast packets 1000
 #
 

配置服务器以主备方式接入。

[~ServerLeaf1_1] 
interface 10GE1/0/2
 port link-type trunk
 undo port trunk allow-pass vlan 1  //不放通VLAN1,防止成环
 storm suppression unknown-unicast 5  //配置未知单播抑制,经验值为10GE端口的5%带宽,建议业务端口都部署
 storm suppression multicast packets 1000  //配置组播报文抑制,接入VXLAN的为组播业务时不可以配置,单播流量可配置,经验值为1000pps
 storm suppression broadcast packets 1000  //配置广播报文抑制,经验值为1000pps,建议业务端口都部署
 stp edged-port enable
#
[~ServerLeaf1_2] 
interface 10GE1/0/2
 port link-type trunk
 undo port trunk allow-pass vlan 1
 storm suppression unknown-unicast 5
 storm suppression multicast packets 1000
 storm suppression broadcast packets 1000
 stp edged-port enable

配置monitor-link关联上行接口和下行接口,避免单台设备的所有上行链路都故障时,本台设备用户侧流量无法转发。
Downlink只列出了1个端口做示例,实际部署时请根据规划补齐。

[~ServerLeaf1_1] 
 monitor-link group 1 
   port 40GE1/0/1 uplink 
   port 40GE1/0/2 uplink 
   port Eth-Trunk10 downlink 1 
   timer recover-time 60   //配置回切时间,防止上行故障回切丢包。
#
[~ServerLeaf1_2] 
 monitor-link group 1 
   port 40GE1/0/1 uplink 
   port 40GE1/0/2 uplink 
   port Eth-Trunk10 downlink 1 
   timer recover-time 60      
 #

配置Underlay路由(以OSPF为例)。

[~ServerLeaf1_1] 
 bfd            //全局使能BFD功能
 #
 ospf 1 router-id 10.125.98.3 
  bfd all-interfaces enable
  bfd all-interfaces min-tx-interval 500 min-rx-interval 500 detect-multiplier 3   //配置BFD的参数为500ms*3
  lsa-arrival-interval intelligent-timer 50 50 50   //设置OSPF LSA接收的时间间隔,优化收敛时间
  area 0.0.0.0 
   network 10.125.97.20 0.0.0.3 
   network 10.125.97.36 0.0.0.3   //分别建立与2台Border Leaf设备的路由邻居
   network 10.125.98.3 0.0.0.0 
   network 10.125.99.2 0.0.0.0   //发布Loopback地址;请勿发布V3版本中作为VXLAN Bypass隧道的Loopback地址
 # 
[~ServerLeaf1_2] 
 bfd            
 #
 ospf 1 router-id 10.125.98.4 
  bfd all-interfaces enable
  bfd all-interfaces min-tx-interval 500 min-rx-interval 500 detect-multiplier 3   //配置BFD的参数为500ms*3
  lsa-arrival-interval intelligent-timer 50 50 50   //优化三层架构,两台物理设备之间多路ECMP情况的OSPF收敛时间
  area 0.0.0.0 
   network 10.125.97.24 0.0.0.3 
   network 10.125.97.40 0.0.0.3 
   network 10.125.98.4 0.0.0.0 
   network 10.125.99.2 0.0.0.0 
 # 

配置网络故障收敛性能优化

[~ServerLeaf1_1][~ServerLeaf1_2]
 interface 40GE1/0/2   
  ospf peer hold-max-cost timer 300000    //所有Spine和Leaf配置OSPF邻居建立后在本地设备的LSA中保持最大开销值的时间300s,源于240s的M-LAG延迟UP时间(同时overlay路由收敛)+ 60s的设备表项同步时间
#
 interface 40GE1/0/3  
  ospf peer hold-max-cost timer 300000    
#

端口优化配置
批量配置端口CRC检测:在peer-link成员口、与Spine互联的口上必须配置。

[~ServerLeaf_1][~ServerLeaf_2]
port-group group-member 40GE 1/0/0 to 40GE 1/0/3
  port crc-statistics trigger error-down
  trap-threshold crc-statistics 100 interval 10 

//配置备份链路的接口检测CRC错包,在接收到CRC错误报文达到告警阈值时触发接口Error-Down,保证数据传输的正确性。
批量关闭不使用的端口并调大stp cost值。

[~ServerLeaf_1][~ServerLeaf_2]
port-group group-member 10GE 1/0/10 to 10GE 1/0/48
  shutdown
  stp instance 0 cost 10000

配置系统定时保存配置的功能
configuration file auto-save interval 360 delay 60 cpu-limit 60

posted @ 2025-04-25 14:03  朱军杰  阅读(164)  评论(0)    收藏  举报