Kept VRRP script, preemptive delay, VIP unicast details

keepalived utilize VRRP Script technology Call external auxiliary script for resource monitoring , And realize optimization according to the monitoring results First dynamically adjust , So as to realize the high availability function of other applications

VRRP Script To configure

Define scripts

vrrp_script： Custom resource monitoring script ,vrrp The instance returns the value according to the script , Public definition , Can be called by multiple instances , set The meaning lies in vrrp Independent configuration blocks outside the instance , Generally placed on global_defs After setting the block .

Log in and copy

      
      

       
       vrrp_script <SCRIPT_NAME> {
       
       
  script <STRING>|<QUOTED-STRING> # The return value of this script is non 0 when , Will trigger the following OPTIONS perform 
       
       
  OPTIONS
       
       
}
      
      
      
      

       
       
1.
       
       
2.
       
       
3.
       
       
4.
      
      

Usually, this script is used to monitor the status of the specified application . Once the application status is abnormal , Trigger pair MASTER The weight of the node is reduced To below SLAVE node , So as to achieve VIP Switch to SLAVE node

Call script

track_script： call vrrp_script Define scripts to monitor resources , It's defined in VRRP Within the example , Call the predefined vrrp_script

Log in and copy

      
      

       
       track_script {
       
       
  SCRIPT_NAME_1
       
       
  SCRIPT_NAME_2
       
       
}
      
      
      
      

       
       
1.
       
       
2.
       
       
3.
       
       
4.
      
      

Definition VRRP script

Log in and copy

      
      

       
       vrrp_script <SCRIPT_NAME> { # Define a detection script , stay global_defs  External configuration 
       
       
  script <STRING>|<QUOTED-STRING> #shell Command or script path 
       
       
  interval <INTEGER> # Time interval between , The unit is in seconds , Default 1 second 
       
       
  timeout <INTEGER> # Timeout time 
       
       
  weight <INTEGER:-254..254> # The default is 0, If this value is set to a negative number , When the return value of the above script is non 0 when , Adding this value to the weight of this node can reduce the weight of this node , It indicates that the fall.  If it's a positive number , When the return value of the script is 0, Adding this value to the weight of this node can improve the weight of this node , It indicates that the  rise. Negative values are usually used 
       
       
  fall <INTEGER> # Executing the script failed several times in a row , Convert to failed , It is recommended to set 2 above 
       
       
  rise <INTEGER> # Execute the script successfully several times in a row , Mark the server from failure to success 
       
       
  user USERNAME [GROUPNAME] # The user or group that executes the monitoring script 
       
       
  init_fail # Set the default flag to failure status , After the monitoring is successful, it will be converted to the successful state 
       
       
}
      
      
      
      

       
       
1.
       
       
2.
       
       
3.
       
       
4.
       
       
5.
       
       
6.
       
       
7.
       
       
8.
       
       
9.
       
       
10.
      
      

call VRRP script

Log in and copy

      
      

       
       vrrp_instance VI_1 {
       
       
  …
       
       
  track_script {
       
       
  chk_down
       
       
  }
       
       
}
      
      
      
      

       
       
1.
       
       
2.
       
       
3.
       
       
4.
       
       
5.
       
       
6.
      
      

Preemptive delay mode preempt_delay

Preemptive delay mode , That is, after the host with high priority is restored , Will not immediately grab back VIP, It's a delay （ Default 300s） Grab back VIP

Log in and copy

      
      

       
       preempt_delay # # The specified preemption delay time is #s, Default delay 300s
       
       
 Be careful ： It needs to be different keepalived The server state by BACKUP, And don't enable  vrrp_strict
       
       

       
       
 Example ：
       
       
#ha1 Host configuration 
       
       
vrrp_instance VI_1 {
       
       
  state BACKUP # All for BACKUP
       
       
  interface eth0
       
       
  virtual_router_id 66
       
       
  priority 100 # High priority 
       
       
  advert_int 1
       
       
  preempt_delay 60 # Preemptive delay mode , Default delay 300s
       
       
#ha2 Host configuration 
       
       
vrrp_instance VI_1 {
       
       
  state BACKUP # All for BACKUP
       
       
  interface eth0
       
       
  virtual_router_id 66
       
       
  priority 80 # Low priority 
       
       
  advert_int 1
      
      
      
      

       
       
1.
       
       
2.
       
       
3.
       
       
4.
       
       
5.
       
       
6.
       
       
7.
       
       
8.
       
       
9.
       
       
10.
       
       
11.
       
       
12.
       
       
13.
       
       
14.
       
       
15.
       
       
16.
       
       
17.
       
       
18.
       
       
19.
      
      

VIP Unicast configuration

Log in and copy

      
      

       
        Default keepalived Multicast is used between hosts to announce messages to each other , Will cause network congestion , Can be replaced by unicast , Reduce network traffic 
       
       

       
       
# At all nodes vrrp_instance Set the name of the opposite host in the statement block IP, It is recommended to set it to the address dedicated to the corresponding core jumper network , Instead of using business networks 
       
       
unicast_src_ip <IPADDR> # Specify the source to send unicast IP
       
       
unicast_peer {
       
       
<IPADDR> # Specify the destination host of the other party receiving unicast IP
       
       
......
       
       
}
      
      
      
      

       
       
1.
       
       
2.
       
       
3.
       
       
4.
       
       
5.
       
       
6.
       
       
7.
       
       
8.