Lvs+kept (DR mode) learning notes

1. sketch

In the medium-sized projects of the Internet , A single server is often unable to meet the performance requirements of the business itself , At this time, it will expand in parallel , Spread the load among several servers ( colony ). Generally, clusters are DNS polling ,LVS,nginx Load balancing .

The main purposes of the cluster include the following three points ： High availability 、 Scalable 、 Load balancing .

A complete cluster architecture , You need to include the following three functions ： Load balancing 、 Fault isolation 、 Failed to switch

Load balancing ： According to the set algorithm , Through the load balancer (Director) Forward external requests to servers in each cluster (Real server)

Fault isolation ： When one or more servers in the cluster (Real server) In case of failure or failure to provide services , The cluster can remove them from the forwarding queue , To ensure that user access will not be incorrectly forwarded to servers that cannot be provided (Real server) Handle . And when the server fails (Real server) When it returns to normal , It can rejoin the forwarding queue of the cluster .

Failed to switch ： Failed to switch master (Master) The key is to remove the load balancer (Director) A single point of question , Once the load balancer (Director) failure , Standby machine (Slave) Can replace the master (Master) Load Balancer (Director) Accept user requests , And when the main load balancer (Director) When it comes back to normal , Able to take over user requests again ( Depending on the configuration )

LVS yes Linux Virtual Linux Abbreviation , namely Linux Virtual server , Is a commonly used server cluster system , This function has been widely integrated in Linux The kernel .

Keepalived Is a server detection status software , It can provide fault isolation and failover functions for clusters

ipvsadm It is LVS Management tools , Ability to add , modify , Delete , Check the configuration and forwarding status of the current cluster

In the use of Keepalived The scheme , As long as the system is installed ipvsadm command , Just go through Keepalived The configuration file of can realize all configurations of the cluster .

This time mainly introduces DR Cluster configuration in mode ,DR Direct routing , Load Balancer (Director) Only negative forwarding of user requests , The request response is provided by the server (Real server) Provide directly .

The following is a simple architecture diagram

2. install

2.1.ipvsadm install

yum install ipvsadm

2.2.Keepalived install

yum install keepalived

2.3.LVS Support view

uname -r
#  see linux Kernel version ,2.4 It has been integrated ,2.4 Please recompile the kernel before ( Now there is no 2.4 Before )

3. To configure

Preset environment ：

Load Balancer A:

        ip Address : 192.168.1.12

         colony VIP:192.168.1.100

Real server B:

        ip Address :192.168.1.13

         Multicast address :192.168.1.100

Real server C:

        ip Address :192.168.1.14

         Multicast address :192.168.1.100

Real server configuration :

Real servers do not need to install any software , Just configure it like an ordinary non clustered server

#  You can write the following commands shell Script , And in /etc/rc.local It is configured to execute automatically every time you start up 
#  Be careful : Must be bound to the loopback network card (lo)
# down To make the script repeatable 
ifconfig lo:0 192.168.1.100 broadcast 192.168.1.100 netmask 255.255.255.255 down
ifconfig lo:0 192.168.1.100 broadcast 192.168.1.100 netmask 255.255.255.255 up
#  This step is very important , In practical applications, multiple real servers in the cluster use the same multicast address , Therefore, it is necessary to suppress ARP radio broadcast 
echo "1" >/proc/sys/net/ipv4/conf/lo/arp_ignore
echo "2" >/proc/sys/net/ipv4/conf/lo/arp_announce
echo "1" >/proc/sys/net/ipv4/conf/all/arp_ignore
echo "2" >/proc/sys/net/ipv4/conf/all/arp_announce

### Please note that ： The physical machine / The virtual machine lo Return address , broadcast, netmask Wait to compare ！！！

After successful execution , Input ifconfig You should see the following results :

lo:0: flags=73<UP,LOOPBACK,RUNNING>  mtu 65536
        inet 192.168.1.100  netmask 255.255.255.255
        loop  txqueuelen 0  (Local Loopback)

Because of inhibition ARP radio broadcast , Input ping 192.168.1.100 It should be the following result :

PING 192.168.1.100 (192.168.1.100) 56(84) bytes of data.
--- 192.168.1.100 ping statistics ---
6 packets transmitted, 0 received, 100% packet loss, time 5042ms

unable ping through

Load balancer configuration

Keepalived Default profile location /etc/keepalived/keepalived.conf

The configuration is as follows :

virtual_server 192.168.1.100 80 {
    delay_loop 6
    lb_algo rr # rr Polling 
    lb_kind DR # DR Pattern 
    nat_mask 255.255.255.255
    #persistence_timeout 600 # Session holding time 
    protocol TCP

    real_server 192.168.1.13 80 {
        #weight 100 # stay wlc Load balancing algorithm takes effect 
        #  Real server status check 
        TCP_CHECK {
            connect_timeout 3
            nb_get_retry 3
            delay_before_retry 3
            connect_port 80
        }
    }

    real_server 192.168.1.14 80 {
        #weight 100 # stay wlc Load balancing algorithm takes effect 
        #  Real server status check 
        TCP_CHECK {
            connect_timeout 3
            nb_get_retry 3
            delay_before_retry 3
            connect_port 80
        }
    }
}

Input keepalived -D Command to start keeplived

see keepalived Running state ( The normal startup is three processes )

ps aux | grep "keepalived"

see keepalived Run log

tail /var/log/message

Check the configuration results ( Under normal circumstances, you can see that the configured information has been loaded correctly )

ipvsadm -L