1, Background

This paper mainly introduces the implementation of presto with keepalive

Experimental environment: CentOS 6 64 bit

 

2, Experimental steps

1. Software installation

Install keepalive package

sudo yum install -y keepalived

presto deployment and configuration are omitted, assuming that the process has been started and port monitoring is in 8083.

2. Write Presto master service survival detection script (required for both machines)
 

$sudo vim /usr/bin/check_presto_alive.sh

#!/bin/sh

PATH=/bin:/sbin:/usr/bin:/usr/sbin

port_test=`nc -z -v  localhost 8083|grep succeeded -c`;

if [ $port_test -eq 0 ]
   then
     echo 'presto server is died'
     killall keepalived
fi

$sudo chmod +x /usr/bin/check_presto_alive.sh

2. Configuration of Presto-master1

$sudo vim /etc/keepalived/keepalived.conf

! Configuration File for keepalived

vrrp_script check_presto_alive {
    script "/usr/bin/check_presto_alive.sh"
    interval 3
    weight -10
}


global_defs {
	router_id LVS_PRESTO #An identification of the machine running keepalived
}

vrrp_instance VI_1 {
	interface eth0 #Set the network card bound by the instance
	state MASTER  #Specify which is the master and which is the backup
	virtual_router_id 92 #VPID flag, the primary and secondary must be the same
	priority 180 #Priority, high priority campaign as master
	vrrp_unicast_bind 192.168.0.1
	vrrp_unicast_peer 192.168.0.2
	authentication {
		auth_type PASS  #Authentication method
		auth_pass nenad #Authentication password
	}
	virtual_ipaddress {
        ## Set VIP, must be virtual IP of the same network segment
        192.168.0.251
	}
    track_script {
        check_presto_alive #presto survival test
    }

}

virtual_server 192.168.0.251 8083 {
    delay_loop 5  #Interval of health examination
    lb_algo rr #LVS scheduling algorithm rr|wrr|lc|wlc|lblc|sh|dh
    lb_kind NAT #Load balancing forwarding rule NAT|DR|RUN
    persistence_timeout 5 #Session hold time
    protocol TCP #Protocol used

    ##Real IP address of RS
    real_server 192.168.0.1 8083 {
        weight 1 #Default is 1, 0 is invalid
        TCP_CHECK{
            connect_timeout 3 #Connection timeout
            nb_get_retry 3 #Reconnection number
            delay_before_retry 3 #Reconnection interval
            connect_port 8083 #Health check port
        }
    }
}

3. Configuration of Presto-master2

$sudo vim /etc/keepalived/keepalived.conf

​
! Configuration File for keepalived

vrrp_script check_presto_alive {
    script "/usr/bin/check_presto_alive.sh"
    interval 3
    weight -10
}


global_defs {
	router_id LVS_PRESTO #An identification of the machine running keepalived
}

vrrp_instance VI_1 {
	interface eth0 #Set the network card bound by the instance
	state BACKUP  #Specify which is the master and which is the backup
	virtual_router_id 92 #VPID flag, the primary and secondary must be the same
	priority 170 #Priority, high priority campaign as master
	vrrp_unicast_bind 192.168.0.2
	vrrp_unicast_peer 192.168.0.1
	authentication {
		auth_type PASS  #Authentication method
		auth_pass nenad #Authentication password
	}
	virtual_ipaddress {
        ## Set VIP, must be virtual IP of the same network segment
        192.168.0.251
	}
    track_script {
        check_presto_alive #presto survival test
    }

}

virtual_server 192.168.0.251 8083 {
    delay_loop 5  #Interval of health examination
    lb_algo rr #LVS scheduling algorithm rr|wrr|lc|wlc|lblc|sh|dh
    lb_kind NAT #Load balancing forwarding rule NAT|DR|RUN
    persistence_timeout 5 #Session hold time
    protocol TCP #Protocol used

    ##Real IP address of RS
    real_server 192.168.0.2 8083 {
        weight 1 #Default is 1, 0 is invalid
        TCP_CHECK{
            connect_timeout 3 #Connection timeout
            nb_get_retry 3 #Reconnection number
            delay_before_retry 3 #Reconnection interval
            connect_port 8083 #Health check port
        }
    }
}

4. Restart keepalive to take effect (both machines execute)

$sudo /etc/init.d/keepalived restart

5. validation

Use the following command to view that the VIP has been bound to a specific network card.

$ ip a

This experiment verifies the automatic drift of VIP and realizes the automatic switch between the master and the backup of Presto master

Note: after repairing the failed service, you must restart the keepalive service of the machine, otherwise keepalive cannot perceive the service recovery!