Reading tips

In this article VIP,vip,Vip It all refers to one thing
Virtual IP—— fictitious IP Address

Environmental Science

1. centos 7.9
2. master server _ Act as the master node (node1):172.24.8.124/16
   Standby server _ Act as a standby node (node2):172.24.8.125/16
3. Virtual IP Set to 172.24.8.242

This configuration is correct under normal operating environment

Master configuration

#yum install keepalived // Installed heartbeat detection package 
#vim /etc/keepalived/keepalived.conf

! Configuration File for keepalived
 
vrrp_script chk_nfs {
    
script "/etc/keepalived/chk_nfs.sh"
interval 2
weight -40
}
vrrp_instance node1 {
    
state MASTER #  Master node flag ——MASTER
interface ens32
virtual_router_id 51
priority 100 #  The highest priority level is 255, The priority of the primary node must be higher than that of the standby node ,
advert_int 1
unicast_src_ip 172.24.8.124
           unicast_peer {
    
		172.24.8.125
           }

authentication {
    
auth_type PASS
auth_pass 1111
}
track_script {
    
chk_nfs
}
notify_stop /etc/keepalived/notify_stop.sh       
notify_master /etc/keepalived/notify_master.sh   
notify_backup /etc/keepalived/notify_backup.sh
virtual_ipaddress {
    
172.24.8.242
}
}

Standby node configuration

#yum install keepalived // Installed heartbeat detection package 
#vim /etc/keepalived/keepalived.conf

! Configuration File for keepalived
 
vrrp_script chk_nfs {
    
script "/etc/keepalived/chk_nfs.sh"
interval 2
weight -40
}
vrrp_instance node2 {
    
state BACKUP  #  Standby node flag ——BACKUP
interface ens32
virtual_router_id 51
priority 99 # The standby node should be smaller than the primary node 
advert_int 1
unicast_src_ip 172.24.8.125 # Local network card IP
           unicast_peer {
    
                172.24.8.124 # Peer network card IP
           }
authentication {
    
auth_type PASS
auth_pass 1111
}
track_script {
    
chk_nfs
}
notify_stop /etc/keepalived/notify_stop.sh       
notify_master /etc/keepalived/notify_master.sh   
notify_backup /etc/keepalived/notify_backup.sh
virtual_ipaddress {
    
172.24.8.242
}
}

Problems found in the test environment

The primary and standby nodes have Virtual IP, Therefore, it needs further analysis and solution .

Analyze and solve

Reasons for appearance

The reason for this problem is that in the server network environment , Because the routing switching layer is disabled ARP The broadcasting restrictions of , cause KEEPALIVE The master / standby protocol can't communicate through broadcast , Both the active and standby servers preempt Virtual-IP Address , Therefore, two servers have the same VIP The situation of address appears .

resolvent

-   Specify... By configuration IP Thin communication between the two servers （ Based on the explanation in Alibaba cloud's description document, only two computers can use the same HAVIP Address ）
-  According to the analysis of the feasibility of this solution , Using the same method for multiple standby should be equally feasible .

stay node1 and node2 Add the following configurations respectively

node1 Configuration added by the master node

    unicast_src_ip  172.24.8.124##（ Local IP Address ）
    unicast_peer {
    
                  172.168.8.125#( Opposite end IP Address ） The address must not be forgotten 
                       }

node2 Configuration added by the standby node

    unicast_src_ip  172.24.8.125##（ Local IP Address ）
    unicast_peer {
    
                  172.168.8.124#( Opposite end IP Address ） The address must not be forgotten 
                       }

By adding unicast_src_ip and unicast_peer Parameter realizes unicast communication between the primary and standby nodes .

Test sequence

1. node1,node2 Enable at the same time keepalived, And check their VIP Address

$ systemctl start keepalived
$ ip a

node1 There is vip

node2 It didn't show up vip

2. node1 close keepalived,Virtual ip towards node2 drift

$ systemctl stop keepalived #node1 operation 
$ ip a #node1,node2 All operations to view vip The change of 

node1 lost vip

node2 To obtain the vip

3. Turn it on again node1 Of keepalived service

$ systemctl start keepalived #node1 operation 
$ ip a #node1,node2 All operations to view vip Address change 

Master node node1 Again Vip

For the node node2 lost Vip