Day 27th: HA High Availability Cluster setup

Small Q: All things must be seen farther, farther, when you take the first step, the heart must be a third step,

The second step can be your chips, no chips of life,

Destined to live in the palm of others.

Today I learned a very good knowledge, I found that my recent slow progress, extenuating, all day just to see two or three knowledge points, but every knowledge point I will study for a long time, although I know I will forget, but

Perhaps I review again, the principle I will not forget, the purpose of the first time to achieve;

Introduction

HA (High Available), highly available cluster, also known as dual standby hot (active/passive) for critical business, is an effective solution to ensure business continuity, usually has two or more than two nodes; when there is a problem with the active node, the standby node is detected at this time. And immediately follow the active node to execute the business. Commonly used high-availability open source software has heartbeat and keepalived, where keepalived has load-balanced functionality.

650) this.width=650; "src=" Http://s3.51cto.com/wyfs02/M00/74/58/wKiom1YaV3XSvqpyAAKWDADGuFc295.jpg "style=" width : 550px;height:314px;float:left; "title=" 11.PNG "width=" 550 "height=" 314 "border=" 0 "hspace=" 0 "vspace=" 0 "alt=" Wkiom1yav3xsvqpyaakwdadgufc295.jpg "/>

650) this.width=650; "src=" Http://s3.51cto.com/wyfs02/M00/74/55/wKioL1YaV5HwbzvpAAFbecF-JuE315.jpg "style=" width : 600px;height:300px; "title=" 111.PNG "width=" "height=" "border=" 0 "hspace=" 0 "vspace=" 0 "alt=" Wkiol1yav5hwbzvpaafbecf-jue315.jpg "/>

Figure one is the way of high-availability clusters, very clear, I do not introduce;

Figure Two is his working principle, explain, just my understanding ah, the wrong words, hope that many advice:

WEB1 server as the main node, WEB2 as the primary standby server, of course, can also add web3 for the main standby service, mobile IP is the virtual IP (VIP), the upper left corner for the user, multicast for the transmission of heartbeat information;

The user via switch to connect the mobile IP (mobile IP at the same time in two servers), the service provided by the WEB1 master server, when the Web1 service is interrupted, WEB2 will detect the heartbeat information to stop, so take the mobile IP, continue the user process

Of course, this process is transparent to the user, she just has been connected to the mobile IP, unless two machines are down, VIP interrupt

Otherwise web1 outage, VIP transfer to WEB2 continue to provide services;

Today's practice is to build high-availability servers with heartbeat.

Pre-preparation

Two servers, I use two virtual machines, IP is the main 192.168.1.103, standby 192.168.1.113, mobile IP 192.168.1.104

For easy experimentation, change the host name

Master: hostname Master >>> Bash

From: hostname slave >>> bash

Dual-machine co-operation

Off firewall: iptables-f setenforce 0 (Getenforce)

Domain point: vim/etc/hosts

192.168.1.103 Master IP host name

192.168.1.113 slave

Install the extension source in order to download the open source software used to configure HA Heartbeat

wget www.lishiming.net/data/attachment/forum/epel-release-6-8_32.noarch.rpm CENTOS6 version 32-bit

Cd/usr/locla/src

RPM-IVH epel-release-6-8_32.noarch.rpm

Yum install-y heartbeat application software for installing HA

Yum install-y libnet ha application-dependent packages

So the prep work is done.

Configure the primary and standby dual machine

Because HA high-availability cluster is only for dual-machine hot standby, so he needs to provide hot standby services, such as MySQL master-slave data hot standby Tomcat Service · Apache Services

Today we use Nginx service to show

Enter the location of the heartbeat newspaper, copy three template files

cd/usr/share/doc/heartbeat-3.0.4/

CP ha.cf Haresources authkeys/etc/ha.d/

cd/etc/ha.d/

VIM Authkeys certification file, about security level settings  

Modify the place as:

650) this.width=650; "src=" Http://s3.51cto.com/wyfs02/M02/74/58/wKiom1YaXHuA_vkNAAB3aEpZ0Ic302.jpg "title=" 1.1. PNG "alt=" Wkiom1yaxhua_vknaab3aepz0ic302.jpg "/>

chmod authkeys Change Permissions

Vim haresources file, used to set the host, mobile IP, hot standby service

Near the bottom of the location add format: Host name VIP service

650) this.width=650; "src=" http://s3.51cto.com/wyfs02/M00/74/58/wKiom1YaXMSSdccVAACklvUz2hE267.jpg "title=" 1.2. PNG "alt=" Wkiom1yaxmssdccvaacklvuz2he267.jpg "/>

192.168.1.114/24/eth0:1

Represents the virtual IP192.168.1 of this network segment, and this IP is eth0 virtual network card, named Eth0:1; add the virtual network card method See me expand the Knowledge module; (The IP in my file is wrong, I'm missing a 1.)

Vim HA.CF This is the most important configuration file, first with redirect empty, add;

Logfile/var/log/ha.log log files

Debugfile/var/log/ha-debug.log

#bcast eth1 Heartbeat using Ethernet broadcast, ETH1 for broadcast address

Ucast eth1 192.168.1.113 Unicast form of communication, with two selected one; NIC information

KeepAlive 2 broadcasts every 2 seconds

Warntime 10 warning time, no signal received from the other node in 10s, send warning to log

Deadtime 30s seizure to the other side of the node signal that the other side down, immediately take over the other node's services

Initdead 120 wait time to restart the system to prevent the system from restarting is not down

Udpport 694 port used for broadcast communication

Auto_failback on if it is on, the service automatically switches back when the primary node fails to recover

Node Master Master host name

node slave slave host name

Ping 192.168.1.11 Quorum node, is a relatively robust device, such as switches, to the same network segment

Respawn Hacluster/usr/lib/heartbeat/ipfail When Heartbeat is started, it is placed in a process that starts with the

it , Ipfail is a tool used to detect network connectivity,

Hacluster is the user who started the program

Then copy three files to the slave

SCP ha.cf haresources Authkeys [email protected]:/etc/ha.d/

Enter the slave password to

Other two files do not change, just change the ha.cf within the ucast, according to your host network card information Change OH

Ucast eth0 192.168.1.103 Note my NIC is eth0

This time the configuration is complete, because we use the Nginx service to test, so we have to install Nginx on both machines

Yum install-y Nginx

650) this.width=650; "src=" Http://s3.51cto.com/wyfs02/M01/74/58/wKiom1YaXn7gDhrkAADSTvlZBok348.jpg "title=" 111111111111.PNG "alt=" Wkiom1yaxn7gdhrkaadstvlzbok348.jpg "/>

Test

Testing is nothing more than to take down the host, to see if the standby server will immediately take over, restore the host, the user process will not automatically switch back;

In fact, is to verify the configuration file configuration can be implemented, of course, some time tired will not be tested;

Start heartbeat, pay special attention to start the sequence, must first start the host

/etc/init.d/heartbeat start

Wait a moment to view the host

Ifconfig will find a virtual NIC read eth0:1

PS aux |grep nginx process is running up.

The same view from the opportunity to find nothing, because it is not possible for both servers to provide services to users

Modify the default page of the Nginx Service Site directory;

(Find the Site directory path into/etc/nginx/nginx.conf >>> include path >>> path/virtual (default). conf >>&G   T Go to find

echo "1111111111master" >/usr/share/nginx

Access 192.168.1.104 Mobile IP

will be displayed: 1111111111master

and the dual-machine hot standby is mainly to detect the user and the server connectivity, when not connected, from the machine to replace him with the user connection;

And the nature of connectivity is that because of the ICMP protocol, shutting it down can disconnect

Iptables-a input-p icmp-j DROP

Host View:

Ifconfig PS aux |grep nginx tail/var/log/ha_log

From Machine view:

Ifconfig PS aux |grep nginx tail/var/log/ha_log

Host Open ICMP protocol

iptables-d input-p icmp-j drop-d Delete

And on the same way to view, will find the host and the user's services automatically switched back;

You can also stop the host's Heartbeat service Test View:

/etc/init.d/heartbeat stop

In fact, the configuration of a variety of highly available services, just changed the service name in the ha.cf file;

Then install the appropriate service software, such as reference to the following MySQL high availability, is a bit too messy

Extensions: concept http://www.linuxidc.com/Linux/2013-08/88522.htm

Select software According to http://svsky.blog.51cto.com/10184644/1672431

MySQL High availability http://tianshi10cool.blog.51cto.com/1870184/1665900

= = = Clearer http://colderboy.blog.51cto.com/485582/104872

Day 27th: HA High Availability Cluster setup