MySQL-MHA集群部署（binlog复制）

MHA的理论知识网上有很多教程，这里不会说明；仅推荐博客链接！

MHA的理论说明：http://www.ywnds.com/?p=8094

MHA的安装包需要在google上面下载，或者就是csdn上面花钱下载！

详细说明怎么搭建MHA

#四台服务器分配如下
10.0.102.214         test3            MHA的管理节点
10.0.102.204         test2            master节点
10.0.102.179         test1            slave节点(作为备用的管理节点)
10.0.102.221         mgt01            slave节点

#这里我们一主两从的架构基于binlog复制，首先需要配置好一主两从的架构。
#需要注意的是，作为备用主的slave服务器需要开通二进制日志和配置log_slave_updates参数
#MySQL基于binglog复制过程如下： https://www.cnblogs.com/wxzhe/p/10051114.html
#部署过程中不会说明怎么搭建MySQL主从架构

第一步：搭建好主从架构，也就是一主两从的架构。【MHA的官方不支持一主一从，但是传闻阿里修改了源码使其支持一主一从，这里使用官方的结构】

需要注意的是要在作为备用主的服务器添加如下配置：

log-bin=                        #开启二进制日志
log_slave_updates               #把SQL线程的动作写入二进制日志

第二步：安装MHA

在MHA的集群的所有服务器上需要安装MHA-node节点，

[root@mgt01 ~]# yum install epel-release perl-DBD-MySQL perl-CPAN -y             #安装依赖包
[root@mgt01 src]# ls
mha4mysql-node-0.56.tar.gz
[root@mgt01 src]# tar zxvf mha4mysql-node-0.56.tar.gz    -C  ../                 #解压
[root@mgt01 src]# cd ../
[root@mgt01 local]# cd mha4mysql-node-0.56/
[root@mgt01 mha4mysql-node-0.56]# ls
AUTHORS  bin  COPYING  debian  inc  lib  Makefile.PL  MANIFEST  META.yml  README  rpm  t
[root@mgt01 mha4mysql-node-0.56]# perl Makefile.PL                               #编译
[root@mgt01 mha4mysql-node-0.56]# make  & make install                           #安装

[root@mgt01 ~]# cd /usr/local/bin                                                #安装完成之后，会在/usr/local/bin目录下面生成如下文件
[root@mgt01 bin]# ls
apply_diff_relay_logs  filter_mysqlbinlog  purge_relay_logs  save_binary_logs
[root@mgt01 bin]# ll
total 44
-r-xr-xr-x 1 root root 16367 Dec  8 10:29 apply_diff_relay_logs
-r-xr-xr-x 1 root root  4807 Dec  8 10:29 filter_mysqlbinlog
-r-xr-xr-x 1 root root  8261 Dec  8 10:29 purge_relay_logs
-r-xr-xr-x 1 root root  7525 Dec  8 10:29 save_binary_logs 

注意上面的这一步操作，需要在MHA集群的每个节点上都执行！

安装MHA-manager，也就是MHA集群的管理节点！

#首先安装MHA-manager需要安装的包
yum install perl-DBD-MySQL perl-Config-Tiny perl-Log-Dispatch perl-Parallel-ForkManager perl-Time-HiRes -y 
#安装MHA-manager
tar zxvf mha4mysql-manager-0.56.tar.gz
cd mha4mysql-manager-0.56/
perl Makefile.PL
make & make install
cp -frp samples/scripts/* /usr/local/bin        #把这些脚本文件拷贝到/usr/local/bin下面，这样不用再添加环境变量

master_ip_failover：故障自动切换时对vip管理的脚本，不是必须。如果我们使用keepalived的，我们可以自己编写脚本完成对vip的管理，比如监控mysql，如果mysql异常，我们停止keepalived就行，这样vip就会自动漂移。

master_ip_online_change：在线切换时对vip的管理，不是必须，同样可以自行编写简单的shell完成。

power_manager：故障发生后关闭主机的脚本，不是必须。

send_report：因故障切换后发送报警的脚本，不是必须，可自行编写简单的shell完成。

脚本说明

第三步：配置MHA

配置MHA这一步主要做的就是写MHA的配置文件，然后创建对应的目录！

在上面的samples/  目录下还有一个目录conf，里面有两个配置文件模板：

[root@test3 conf]# ls
app1.cnf  masterha_default.cnf
[root@test3 conf]# 

把配置文件模板拷贝到/etc下面：

mkdir /etc/masterha -p             #在/etc下面创建MHA使用的配置文件的目录【名字可以随意，最好可以标识目录的内容】
cp * /etc/masterha/

首先编辑masterha_default.cnf文件：

[root@mha ~]# cat /etc/masterha_default.cnf
[server default]
# 设置监控用户mha,需要有授权
user=mha
# 设置mysql中root用户的密码,这个密码是前文中创建监控用户的那个密码;
password=
# 设置复制环境中的复制用户名;
repl_user=repl
# 设置复制用户的密码;
repl_password=
# 设置ssh的登录用户名;
ssh_user=root
# 设置ssh的登录端口（不写默认22端口）;
ssh_port=
# 设置监控主库,发送ping包的时间间隔,默认是3秒,尝试三次没有回应的时候自动进行failover;
ping_interval=
# 设置mysql master保存binlog的目录,以便MHA可以找到master的二进制日志;
master_binlog_dir= /data/mysql/
# 设置mysql master在发生切换时保存binlog的目录,在mysql master上创建这个目录（不写默认为/var/tmp）;
remote_workdir=/data/log/masterha

# 一旦MHA到mysql01的监控之间出现问题,MHA Manager将会尝试从mysql02,mysql03登录到mysql01;
secondary_check_script= masterha_secondary_check -s test1 -s mgt01 --user=root --port= --master_host=test2 --master_port=
# 设置自动failover时候的切换脚本(脚本有瑕疵,需要自行修改);
#master_ip_failover_script=/usr/local/bin/master_ip_failover
# 设置手动切换时候的切换脚本（脚本有瑕疵,需要自行修改）;
#master_ip_online_change_script=/usr/local/bin/master_ip_online_change
# 设置发生切换后发送的报警的脚本（可自行编写）;
#report_script=/usr/local/bin/send_report
# 设置故障发生后关闭故障主机脚本（该脚本的主要作用是关闭主机放在发生脑裂,这里没有使用）;
#shutdown_script=""

masterha_default.cnf配置参数说明

上面给出了masterha_default.cnf每个配置参数的说明情况，下面这个是我的配置

[server default]
user=root
password=
ssh_user=root
ssh_port=
ping_interval=
repl_user=repl
repl_password=

master_binlog_dir= /data/mysql/

remote_workdir=/data/log/masterha

secondary_check_script= masterha_secondary_check -s test1 -s mgt01 --user=root --port= --master_host=test2 --master_port=

master_ip_failover_script= /usr/local/bin/master_ip_failover
# shutdown_script= /script/masterha/power_manager
report_script= /usr/local/bin/send_report
# master_ip_online_change_script= /script/masterha/master_ip_online_change

然后编辑配置app1.conf文件

只针对单个应用生效，但是app1.cnf的配置参数优先级高于masterha_default.cnf，一般都会在app1.cnf包含masterha_default.cnf所有参数。MHA可以监控多个主从的集群，每个集群的配置文件可以用名字区分，因为这里只有一个集群，因此只有app1.conf一个文件！

[root@test3 masterha]# cat app1.cnf

manager_log=/data/log/app1/manager.log
manager_workdir=/data/log/app1
master_binlog_dir=/data/mysql
password=123456
ping_interval=3
remote_workdir=/data/log/masterha
repl_password=123456
repl_user=repl
report_script=/usr/local/bin/send_report
secondary_check_script=masterha_secondary_check -s test1 -s mgt01 --user=root --port=22 --master_host=test2 --master_port=3306
ssh_port=22
ssh_user=root
user=root

[server1]
hostname=10.0.102.204
port=3306
candidate_master=1

[server2]
candidate_master=1
hostname=10.0.102.179
port=3306

[server3]
hostname=10.0.102.221
no_master=1
port=3306

这个配置文件的参数基本都比较好理解，需要注意的是，配置文件指定的目录都需要另行创建！

mkdir -p /data/log/masterha
mkdir /data/log/app1

当candidate_master设置为1时，表示为候选master，如果设置该参数以后，发生主从切换以后将会将此从库提升为主库，即使这个主库不是集群中事件最新的slave。默认情况下如果一个slave落后master 100M的relay logs的话，MHA将不会选择该slave作为一个新的master，因为对于这个slave的恢复需要花费很长时间，通过设置check_repl_delay=0，MHA触发切换在选择一个新的master的时候将会忽略复制延时，这个参数对于设置了candidate_master=1的主机非常有用，因为这个候选主在切换的过程中一定是新的master check_repl_delay=0。

同样设置为候选master的slave一定要开启二进制日志和log_slave_updates参数！

设置relay log的清除方式（在每个slave节点上）

在配置文件中加上relay_log_purge=0,需要重启才能生效！

注意：MHA在发生切换的过程中，从库的恢复过程中依赖于relay log的相关信息，所以这里要将relay log的自动清除设置为OFF，采用手动清除relay log的方式。在默认情况下，从服务器上的中继日志会在SQL线程执行完毕后被自动删除。但是在MHA环境中，这些中继日志在恢复其他从服务器时可能会被用到，因此需要禁用中继日志的自动删除功能。定期清除中继日志需要考虑到复制延时的问题。在ext3的文件系统下，删除大的文件需要一定的时间，会导致严重的复制延时。为了避免复制延时，需要暂时为中继日志创建硬链接，因为在Linux系统中通过硬链接删除大文件速度会很快。（在mysql数据库中，删除大表时，通常也采用建立硬链接的方式）

MHA节点中包含了pure_relay_logs命令工具，它可以为中继日志创建硬链接，执行SET GLOBAL relay_log_purge=1，等待几秒钟以便SQL线程切换到新的中继日志，再执行SET GLOBAL relay_log_purge=0。

pure_relay_logs脚本参数如下所示：

--user mysql                 #用户名;
--password mysql             #密码;
--port                       #端口号;
--workdir                    #指定创建relay log的硬链接的位置，默认是/var/tmp，由于系统不同分区创建硬链接文件会失败，故需要执行硬链接具体位置，成功执行脚本后，硬链接的中继日志文件被删除;
--disable_relay_log_purge    #默认情况下，如果relay_log_purge=，脚本会什么都不清理，自动退出，通过设定这个参数，当relay_log_purge=1的情况下会将relay_log_purge设置为0。清理relay log之后，最后将参数设置为OFF;

设置定期清理relay脚本

[root@mgt01 ~]# cat !$
cat purge_relay.sh
#!/bin/bash
user=root
passwd=
port=
log_dir='/data/masterha/log'
work_dir='/data'
purge='/usr/local/bin/purge_relay_logs'

if [ ! -d $log_dir ];then
  mkdir $log_dir -p
fi

$purge --user=$user --password=$passwd --disable_relay_log_purge --port=$port --workdir=$work_dir >> $log_dir/purge_relay_logs.log >&

把以上脚本加入定时计划任务：

[root@mgt01 log]# crontab -l
*  * * * sh /root/purge_relay.sh

purge_relay_logs脚本删除中继日志不会阻塞SQL线程

第四步： 设置ssh无密码认证

MHA的管理节点可以无密码访问集群中的其余节点！

MySQL集群需要互相之间可以无密码访问！

ssh无密码访问不再写过程。

使用MHA检查ssh是否成功

[root@test3 ~]# masterha_check_ssh --conf=/etc/masterha/app1.cnf

若成功则进行下一步，检查复制

【有一些博客提到：暂时先注释配置文件中master_ip_failover_script= /usr/local/bin/master_ip_failover这个选项，不然这个检查过不去的。但是我测试时候没有注释，也是可以检查成功的】

[root@test3 ~]# masterha_check_repl --conf=/etc/masterha/app1.cnf

[root@test3 ~]# masterha_check_repl --conf=/etc/masterha/app1.cnf
Sat Dec   ::  - [warning] Global configuration file /etc/masterha_default.cnf not found. Skipping.
Sat Dec   ::  - [info] Reading application default configuration from /etc/masterha/app1.cnf..
Sat Dec   ::  - [info] Reading server configuration from /etc/masterha/app1.cnf..
Sat Dec   ::  - [info] MHA::MasterMonitor version 0.56.
Sat Dec   ::  - [info] GTID failover mode =
Sat Dec   ::  - [info] Dead Servers:
Sat Dec   ::  - [info] Alive Servers:
Sat Dec   ::  - [info]   10.0.102.204(10.0.102.204:)
Sat Dec   ::  - [info]   10.0.102.179(10.0.102.179:)
Sat Dec   ::  - [info]   10.0.102.221(10.0.102.221:)
Sat Dec   ::  - [info] Alive Slaves:
Sat Dec   ::  - [info]   10.0.102.179(10.0.102.179:)  Version=5.7.-log (oldest major version between slaves) log-bin:enabled
Sat Dec   ::  - [info]     Replicating from 10.0.102.204(10.0.102.204:)
Sat Dec   ::  - [info]     Primary candidate for the new Master (candidate_master is set)
Sat Dec   ::  - [info]   10.0.102.221(10.0.102.221:)  Version=5.7. (oldest major version between slaves) log-bin:disabled
Sat Dec   ::  - [info]     Replicating from 10.0.102.204(10.0.102.204:)
Sat Dec   ::  - [info]     Not candidate for the new Master (no_master is set)
Sat Dec   ::  - [info] Current Alive Master: 10.0.102.204(10.0.102.204:)
Sat Dec   ::  - [info] Checking slave configurations..
Sat Dec   ::  - [info]  read_only= is not set on slave 10.0.102.179(10.0.102.179:).
Sat Dec   ::  - [info]  read_only= is not set on slave 10.0.102.221(10.0.102.221:).
Sat Dec   ::  - [warning]  log-bin is not set on slave 10.0.102.221(10.0.102.221:). This host cannot be a master.
Sat Dec   ::  - [info] Checking replication filtering settings..
Sat Dec   ::  - [info]  binlog_do_db= , binlog_ignore_db=
Sat Dec   ::  - [info]  Replication filtering check ok.
Sat Dec   ::  - [info] GTID (with auto-pos) is not supported
Sat Dec   ::  - [info] Starting SSH connection tests..
Sat Dec   ::  - [info] All SSH connection tests passed successfully.
Sat Dec   ::  - [info] Checking MHA Node version..
Sat Dec   ::  - [info]  Version check ok.
Sat Dec   ::  - [info] Checking SSH publickey authentication settings on the current master..
Sat Dec   ::  - [info] HealthCheck: SSH to 10.0.102.204 is reachable.
Sat Dec   ::  - [info] Master MHA Node version is 0.56.
Sat Dec   ::  - [info] Checking recovery script configurations on 10.0.102.204(10.0.102.204:)..
Sat Dec   ::  - [info]   Executing command: save_binary_logs --command=test --start_pos= --binlog_dir=/data/mysql --output_file=/data/log/masterha/save_binary_logs_test --manager_version=0.56 --start_file=test2-bin.
Sat Dec   ::  - [info]   Connecting to root@10.0.102.204(10.0.102.204:)..
  Creating /data/log/masterha if not exists..    ok.
  Checking output directory is accessible or not..
   ok.
  Binlog found at /data/mysql, up to test2-bin.
Sat Dec   ::  - [info] Binlog setting check done.
Sat Dec   ::  - [info] Checking SSH publickey authentication and checking recovery script configurations on all alive slave servers..
Sat Dec   ::  - [info]   Executing command : apply_diff_relay_logs --command=test --slave_user='root' --slave_host=10.0.102.179 --slave_ip=10.0.102.179 --slave_port= --workdir=/data/log/masterha --target_version=5.7.-log --manager_version=0.56 --relay_log_info=/data/mysql/relay-log.info  --relay_dir=/data/mysql/  --slave_pass=xxx
Sat Dec   ::  - [info]   Connecting to root@10.0.102.179(10.0.102.179:)..
  Checking slave recovery environment settings..
    Opening /data/mysql/relay-log.info ... ok.
    Relay log found at /data/mysql, up to test1-relay-bin.
    Temporary relay log file is /data/mysql/test1-relay-bin.
    Testing mysql connection and privileges..mysql: [Warning] Using a password on the command line interface can be insecure.
 done.
    Testing mysqlbinlog output.. done.
    Cleaning up test file(s).. done.
Sat Dec   ::  - [info]   Executing command : apply_diff_relay_logs --command=test --slave_user='root' --slave_host=10.0.102.221 --slave_ip=10.0.102.221 --slave_port= --workdir=/data/log/masterha --target_version=5.7. --manager_version=0.56 --relay_log_info=/data/mysql/relay-log.info  --relay_dir=/data/mysql/  --slave_pass=xxx
Sat Dec   ::  - [info]   Connecting to root@10.0.102.221(10.0.102.221:)..
  Checking slave recovery environment settings..
    Opening /data/mysql/relay-log.info ... ok.
    Relay log found at /data/mysql, up to mgt01-relay-bin.
    Temporary relay log file is /data/mysql/mgt01-relay-bin.
    Testing mysql connection and privileges..mysql: [Warning] Using a password on the command line interface can be insecure.
 done.
    Testing mysqlbinlog output.. done.
    Cleaning up test file(s).. done.
Sat Dec   ::  - [info] Slaves settings check done.
Sat Dec   ::  - [info]
10.0.102.204(10.0.102.204:) (current master)
 +--10.0.102.179(10.0.102.179:)
 +--10.0.102.221(10.0.102.221:)

Sat Dec   ::  - [info] Checking replication health on 10.0.102.179..
Sat Dec   ::  - [info]  ok.
Sat Dec   ::  - [info] Checking replication health on 10.0.102.221..
Sat Dec   ::  - [info]  ok.
Sat Dec   ::  - [warning] master_ip_failover_script is not defined.
Sat Dec   ::  - [warning] shutdown_script is not defined.
Sat Dec   ::  - [info] Got exit code  (Not master dead).

MySQL Replication Health is OK.详细过程

详细过程

遇到过一次是复制检查时，总是会dead servers下面有一个服务器，但是集群里面是正常的，各种都是正常的，后来发现是本地的解析出错！【/etc/hosts文化和ssh目录下面的known_hosts文件，新建的服务器一般不会出现这问题】

查看MHA-manger的状态

[root@test3 masterha]# masterha_check_status --conf=/etc/masterha/app1.cnf
app1 is stopped(:NOT_RUNNING).
[root@test3 masterha]# 

开启MHa-manager

[root@test3 masterha]# nohup masterha_manager --conf=/etc/masterha/app1.cnf --remove_dead_master_conf --ignore_last_failover &

#参数说明
remove_dead_master_conf：设置了这个参数后，如果MHA failover结束后，MHA Manager会自动在配置文件中删除dead master的相关项。如果不设置，
由于dead master的配置还存在文件中，那么当MHA failover后，当再次restart MHA manager后，会报错（there is a dead slave previous dead master）。
ignore_last_failover：默认情况下，如果一个或者多个slave down掉了，master monitor进程就会停掉，就算你设置了ignore_fail。如果设置了–ignore_fail_on_start参数，ignore_fail标记了slave挂掉也不会让master monitor进程停掉。

启动之后查看状态：

[root@test3 masterha]# masterha_check_status --conf=/etc/masterha/app1.cnf
app1 (pid:) is running(:PING_OK), master:10.0.102.204

如果启动没有报错，那么一个MHA的集群就已经搭建成功！

关闭MHA-manager可以使用如下命令：

 masterha_stop --conf=/etc/masterha/app1.cnf

最后：我们进行一个failover测试！

停掉MySQL主从集群中的主，查看是否会自动切换到从！在测试主从之前最后可以写入一点数据，这里我利用tpcc写入了一些数据！

./tpcc_load -h 10.0.102.204 -P  -d tpcc_test -u root -p  -w 3

tpcc的测试使用：https://www.cnblogs.com/wxzhe/p/10027474.html

停掉当前的主服务器！

[root@test2 ~]# service mysqld stop
Shutting down MySQL............ SUCCESS!

然后查看MHA的管理日志

Sat Dec   ::  - [info] Executing secondary network check script: masterha_secondary_check -s test1 -s mgt01 --user=root --port= --master_host=test2 --master_port=  --user=root  --master_host=10.0.102.204  --master_ip=10.0.102.204  --master_port= --master_user=root --master_password= --ping_type=SELECT
Sat Dec   ::  - [info] Executing SSH check script: save_binary_logs --command=test --start_pos= --binlog_dir=/data/mysql --output_file=/data/log/masterha/save_binary_logs_test --manager_version=0.56 --binlog_prefix=test2-bin
Monitoring server test1 is reachable, Master is not reachable from test1. OK.
Sat Dec   ::  - [info] HealthCheck: SSH to 10.0.102.204 is reachable.
Monitoring server mgt01 is reachable, Master is not reachable from mgt01. OK.
Sat Dec   ::  - [info] Master is not reachable from all other monitoring servers. Failover should start.
Sat Dec   ::  - [warning] Got error on MySQL connect:  (Lost connection to MySQL server at 'reading initial communication packet', system error: )
Sat Dec   ::  - [warning] Connection failed  time(s)..
Sat Dec   ::  - [warning] Got error on MySQL connect:  (Lost connection to MySQL server at 'reading initial communication packet', system error: )
Sat Dec   ::  - [warning] Connection failed  time(s)..
Sat Dec   ::  - [warning] Got error on MySQL connect:  (Lost connection to MySQL server at 'reading initial communication packet', system error: )
Sat Dec   ::  - [warning] Connection failed  time(s)..
Sat Dec   ::  - [warning] Master is not reachable from health checker!
Sat Dec   ::  - [warning] Master 10.0.102.204(10.0.102.204:) is not reachable!
Sat Dec   ::  - [warning] SSH is reachable.
Sat Dec   ::  - [info] Connecting to a master server failed. Reading configuration file /etc/masterha_default.cnf and /etc/masterha/app1.cnf again, and trying to connect to all servers to check server status..
Sat Dec   ::  - [warning] Global configuration file /etc/masterha_default.cnf not found. Skipping.
Sat Dec   ::  - [info] Reading application default configuration from /etc/masterha/app1.cnf..
Sat Dec   ::  - [info] Reading server configuration from /etc/masterha/app1.cnf..
Sat Dec   ::  - [info] GTID failover mode =
Sat Dec   ::  - [info] Dead Servers:
Sat Dec   ::  - [info]   10.0.102.204(10.0.102.204:)
Sat Dec   ::  - [info] Alive Servers:
Sat Dec   ::  - [info]   10.0.102.179(10.0.102.179:)
Sat Dec   ::  - [info]   10.0.102.221(10.0.102.221:)
Sat Dec   ::  - [info] Alive Slaves:
Sat Dec   ::  - [info]   10.0.102.179(10.0.102.179:)  Version=5.7.-log (oldest major version between slaves) log-bin:enabled
Sat Dec   ::  - [info]     Replicating from 10.0.102.204(10.0.102.204:)
Sat Dec   ::  - [info]     Primary candidate for the new Master (candidate_master is set)
Sat Dec   ::  - [info]   10.0.102.221(10.0.102.221:)  Version=5.7. (oldest major version between slaves) log-bin:disabled
Sat Dec   ::  - [info]     Replicating from 10.0.102.204(10.0.102.204:)
Sat Dec   ::  - [info]     Not candidate for the new Master (no_master is set)
Sat Dec   ::  - [info] Checking slave configurations..
Sat Dec   ::  - [info]  read_only= is not set on slave 10.0.102.179(10.0.102.179:).
Sat Dec   ::  - [info]  read_only= is not set on slave 10.0.102.221(10.0.102.221:).
Sat Dec   ::  - [warning]  log-bin is not set on slave 10.0.102.221(10.0.102.221:). This host cannot be a master.
Sat Dec   ::  - [info] Checking replication filtering settings..
Sat Dec   ::  - [info]  Replication filtering check ok.
Sat Dec   ::  - [info] Master is down!
Sat Dec   ::  - [info] Terminating monitoring script.
Sat Dec   ::  - [info] Got exit code  (Master dead).
Sat Dec   ::  - [info] MHA::MasterFailover version 0.56.
Sat Dec   ::  - [info] Starting master failover.
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] * Phase : Configuration Check Phase..
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] GTID failover mode =
Sat Dec   ::  - [info] Dead Servers:
Sat Dec   ::  - [info]   10.0.102.204(10.0.102.204:)
Sat Dec   ::  - [info] Checking master reachability via MySQL(double check)...
Sat Dec   ::  - [info]  ok.
Sat Dec   ::  - [info] Alive Servers:
Sat Dec   ::  - [info]   10.0.102.179(10.0.102.179:)
Sat Dec   ::  - [info]   10.0.102.221(10.0.102.221:)
Sat Dec   ::  - [info] Alive Slaves:
Sat Dec   ::  - [info]   10.0.102.179(10.0.102.179:)  Version=5.7.-log (oldest major version between slaves) log-bin:enabled
Sat Dec   ::  - [info]     Replicating from 10.0.102.204(10.0.102.204:)
Sat Dec   ::  - [info]     Primary candidate for the new Master (candidate_master is set)
Sat Dec   ::  - [info]   10.0.102.221(10.0.102.221:)  Version=5.7. (oldest major version between slaves) log-bin:disabled
Sat Dec   ::  - [info]     Replicating from 10.0.102.204(10.0.102.204:)
Sat Dec   ::  - [info]     Not candidate for the new Master (no_master is set)
Sat Dec   ::  - [info] Starting Non-GTID based failover.
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] ** Phase : Configuration Check Phase completed.
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] * Phase : Dead Master Shutdown Phase..
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] Forcing shutdown so that applications never connect to the current master..
Sat Dec   ::  - [warning] master_ip_failover_script is not set. Skipping invalidating dead master IP address.
Sat Dec   ::  - [warning] shutdown_script is not set. Skipping explicit shutting down of the dead master.
Sat Dec   ::  - [info] * Phase : Dead Master Shutdown Phase completed.
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] * Phase : Master Recovery Phase..
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] * Phase 3.1: Getting Latest Slaves Phase..
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] The latest binary log file/position on all slaves is test2-bin.:
Sat Dec   ::  - [info] Latest slaves (Slaves that received relay log files to the latest):
Sat Dec   ::  - [info]   10.0.102.179(10.0.102.179:)  Version=5.7.-log (oldest major version between slaves) log-bin:enabled
Sat Dec   ::  - [info]     Replicating from 10.0.102.204(10.0.102.204:)
Sat Dec   ::  - [info]     Primary candidate for the new Master (candidate_master is set)
Sat Dec   ::  - [info]   10.0.102.221(10.0.102.221:)  Version=5.7. (oldest major version between slaves) log-bin:disabled
Sat Dec   ::  - [info]     Replicating from 10.0.102.204(10.0.102.204:)
Sat Dec   ::  - [info]     Not candidate for the new Master (no_master is set)
Sat Dec   ::  - [info] The oldest binary log file/position on all slaves is test2-bin.:
Sat Dec   ::  - [info] Oldest slaves:
Sat Dec   ::  - [info]   10.0.102.179(10.0.102.179:)  Version=5.7.-log (oldest major version between slaves) log-bin:enabled
Sat Dec   ::  - [info]     Replicating from 10.0.102.204(10.0.102.204:)
Sat Dec   ::  - [info]     Primary candidate for the new Master (candidate_master is set)
Sat Dec   ::  - [info]   10.0.102.221(10.0.102.221:)  Version=5.7. (oldest major version between slaves) log-bin:disabled
Sat Dec   ::  - [info]     Replicating from 10.0.102.204(10.0.102.204:)
Sat Dec   ::  - [info]     Not candidate for the new Master (no_master is set)
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] * Phase 3.2: Saving Dead Master's Binlog Phase..
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] Fetching dead master's binary logs..
Sat Dec   ::  - [info] Executing command on the dead master 10.0.102.204(10.0.102.204:): save_binary_logs --command=save --start_file=test2-bin.  --start_pos= --binlog_dir=/data/mysql --output_file=/data/log/masterha/saved_master_binlog_from_10.0.102.204_3306_20181208172159.binlog --handle_raw_binlog= --disable_log_bin= --manager_version=0.56
  Creating /data/log/masterha if not exists..    ok.
 Concat binary/relay logs from test2-bin. pos  to test2-bin. EOF into /data/log/masterha/saved_master_binlog_from_10.0.102.204_3306_20181208172159.binlog ..
 Binlog Checksum enabled
  Dumping binlog format description event, from position  to .. ok.
  Dumping effective binlog data from /data/mysql/test2-bin. position  to tail().. ok.
 Binlog Checksum enabled
 Concat succeeded.
Sat Dec   ::  - [info] scp from root@10.0.102.204:/data/log/masterha/saved_master_binlog_from_10.0.102.204_3306_20181208172159.binlog to local:/data/log/app1/saved_master_binlog_from_10.0.102.204_3306_20181208172159.binlog succeeded.
Sat Dec   ::  - [info] HealthCheck: SSH to 10.0.102.179 is reachable.
Sat Dec   ::  - [info] HealthCheck: SSH to 10.0.102.221 is reachable.
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] * Phase 3.3: Determining New Master Phase..
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] Finding the latest slave that has all relay logs for recovering other slaves..
Sat Dec   ::  - [info] All slaves received relay logs to the same position. No need to resync each other.
Sat Dec   ::  - [info] Searching new master from slaves..
Sat Dec   ::  - [info]  Candidate masters from the configuration file:
Sat Dec   ::  - [info]   10.0.102.179(10.0.102.179:)  Version=5.7.-log (oldest major version between slaves) log-bin:enabled
Sat Dec   ::  - [info]     Replicating from 10.0.102.204(10.0.102.204:)
Sat Dec   ::  - [info]     Primary candidate for the new Master (candidate_master is set)
Sat Dec   ::  - [info]  Non-candidate masters:
Sat Dec   ::  - [info]   10.0.102.221(10.0.102.221:)  Version=5.7. (oldest major version between slaves) log-bin:disabled
Sat Dec   ::  - [info]     Replicating from 10.0.102.204(10.0.102.204:)
Sat Dec   ::  - [info]     Not candidate for the new Master (no_master is set)
Sat Dec   ::  - [info]  Searching from candidate_master slaves which have received the latest relay log events..
Sat Dec   ::  - [info] New master is 10.0.102.179(10.0.102.179:)
Sat Dec   ::  - [info] Starting master failover..
Sat Dec   ::  - [info]
From:
10.0.102.204(10.0.102.204:) (current master)
 +--10.0.102.179(10.0.102.179:)
 +--10.0.102.221(10.0.102.221:)

To:
10.0.102.179(10.0.102.179:) (new master)
 +--10.0.102.221(10.0.102.221:)
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] * Phase 3.3: New Master Diff Log Generation Phase..
Sat Dec   ::  - [info]
Sat Dec   ::  - [info]  This server has all relay logs. No need to generate diff files from the latest slave.
Sat Dec   ::  - [info] Sending binlog..
Sat Dec   ::  - [info] scp from local:/data/log/app1/saved_master_binlog_from_10.0.102.204_3306_20181208172159.binlog to root@10.0.102.179:/data/log/masterha/saved_master_binlog_from_10.0.102.204_3306_20181208172159.binlog succeeded.
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] * Phase 3.4: Master Log Apply Phase..
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] *NOTICE: If any error happens from this phase, manual recovery is needed.
Sat Dec   ::  - [info] Starting recovery on 10.0.102.179(10.0.102.179:)..
Sat Dec   ::  - [info]  Generating diffs succeeded.
Sat Dec   ::  - [info] Waiting until all relay logs are applied.
Sat Dec   ::  - [info]  done.
Sat Dec   ::  - [info] Getting slave status..
Sat Dec   ::  - [info] This slave(10.0.102.179)'s Exec_Master_Log_Pos equals to Read_Master_Log_Pos(test2-bin.000007:154). No need to recover from Exec_Master_Log_Pos.
Sat Dec   ::  - [info] Connecting to the target slave host 10.0.102.179, running recover script..
Sat Dec   ::  - [info] Executing command: apply_diff_relay_logs --command=apply --slave_user='root' --slave_host=10.0.102.179 --slave_ip=10.0.102.179  --slave_port= --apply_files=/data/log/masterha/saved_master_binlog_from_10.0.102.204_3306_20181208172159.binlog --workdir=/data/log/masterha --target_version=5.7.-log --timestamp= --handle_raw_binlog= --disable_log_bin= --manager_version=0.56 --slave_pass=xxx
Sat Dec   ::  - [info]
MySQL client version is 5.7.. Using --binary-mode.
Applying differential binary/relay log files /data/log/masterha/saved_master_binlog_from_10.0.102.204_3306_20181208172159.binlog on 10.0.102.179:. This may take long time...
Applying log files succeeded.
Sat Dec   ::  - [info]  All relay logs were successfully applied.
Sat Dec   ::  - [info] Getting new master's binlog name and position..
Sat Dec   ::  - [info]  test1-bin.:
Sat Dec   ::  - [info]  All other slaves should start replication from here. Statement should be: CHANGE MASTER TO MASTER_HOST='10.0.102.179', MASTER_PORT=, MASTER_LOG_FILE='test1-bin.000001', MASTER_LOG_POS=, MASTER_USER='repl', MASTER_PASSWORD='xxx';
Sat Dec   ::  - [warning] master_ip_failover_script is not set. Skipping taking over new master IP address.
Sat Dec   ::  - [info] ** Finished master recovery successfully.
Sat Dec   ::  - [info] * Phase : Master Recovery Phase completed.
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] * Phase : Slaves Recovery Phase..
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] * Phase 4.1: Starting Parallel Slave Diff Log Generation Phase..
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] -- Slave diff file generation on host 10.0.102.221(10.0.102.221:) started, pid: . Check tmp log /data/log/app1/10.0..221_3306_20181208172159.log if it takes time..
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] Log messages from 10.0.102.221 ...
Sat Dec   ::  - [info]
Sat Dec   ::  - [info]  This server has all relay logs. No need to generate diff files from the latest slave.
Sat Dec   ::  - [info] End of log messages from 10.0.102.221.
Sat Dec   ::  - [info] -- 10.0.102.221(10.0.102.221:) has the latest relay log events.
Sat Dec   ::  - [info] Generating relay diff files from the latest slave succeeded.
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] * Phase 4.2: Starting Parallel Slave Log Apply Phase..
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] -- Slave recovery on host 10.0.102.221(10.0.102.221:) started, pid: . Check tmp log /data/log/app1/10.0..221_3306_20181208172159.log if it takes time..
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] Log messages from 10.0.102.221 ...
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] Sending binlog..
Sat Dec   ::  - [info] scp from local:/data/log/app1/saved_master_binlog_from_10.0.102.204_3306_20181208172159.binlog to root@10.0.102.221:/data/log/masterha/saved_master_binlog_from_10.0.102.204_3306_20181208172159.binlog succeeded.
Sat Dec   ::  - [info] Starting recovery on 10.0.102.221(10.0.102.221:)..
Sat Dec   ::  - [info]  Generating diffs succeeded.
Sat Dec   ::  - [info] Waiting until all relay logs are applied.
Sat Dec   ::  - [info]  done.
Sat Dec   ::  - [info] Getting slave status..
Sat Dec   ::  - [info] This slave(10.0.102.221)'s Exec_Master_Log_Pos equals to Read_Master_Log_Pos(test2-bin.000007:154). No need to recover from Exec_Master_Log_Pos.
Sat Dec   ::  - [info] Connecting to the target slave host 10.0.102.221, running recover script..
Sat Dec   ::  - [info] Executing command: apply_diff_relay_logs --command=apply --slave_user='root' --slave_host=10.0.102.221 --slave_ip=10.0.102.221  --slave_port= --apply_files=/data/log/masterha/saved_master_binlog_from_10.0.102.204_3306_20181208172159.binlog --workdir=/data/log/masterha --target_version=5.7. --timestamp= --handle_raw_binlog= --disable_log_bin= --manager_version=0.56 --slave_pass=xxx
Sat Dec   ::  - [info]
MySQL client version is 5.7.. Using --binary-mode.
Applying differential binary/relay log files /data/log/masterha/saved_master_binlog_from_10.0.102.204_3306_20181208172159.binlog on 10.0.102.221:. This may take long time...
Applying log files succeeded.
Sat Dec   ::  - [info]  All relay logs were successfully applied.
Sat Dec   ::  - [info]  Resetting slave 10.0.102.221(10.0.102.221:) and starting replication from the new master 10.0.102.179(10.0.102.179:)..
Sat Dec   ::  - [info]  Executed CHANGE MASTER.
Sat Dec   ::  - [info]  Slave started.
Sat Dec   ::  - [info] End of log messages from 10.0.102.221.
Sat Dec   ::  - [info] -- Slave recovery on host 10.0.102.221(10.0.102.221:) succeeded.
Sat Dec   ::  - [info] All new slave servers recovered successfully.
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] * Phase : New master cleanup phase..
Sat Dec   ::  - [info]
Sat Dec   ::  - [info] Resetting slave info on the new master..
Sat Dec   ::  - [info]  10.0.102.179: Resetting slave info succeeded.
Sat Dec   ::  - [info] Master failover to 10.0.102.179(10.0.102.179:) completed successfully.
Sat Dec   ::  - [info] Deleted server1 entry from /etc/masterha/app1.cnf .
Sat Dec   ::  - [info] 

----- Failover Report -----

app1: MySQL Master failover 10.0.102.204(10.0.102.204:) to 10.0.102.179(10.0.102.179:) succeeded

Master 10.0.102.204(10.0.102.204:) is down!

Check MHA Manager logs at test3:/data/log/app1/manager.log for details.

Started automated(non-interactive) failover.
The latest slave 10.0.102.179(10.0.102.179:) has all relay logs for recovery.
Selected 10.0.102.179(10.0.102.179:) as a new master.
10.0.102.179(10.0.102.179:): OK: Applying all logs succeeded.
10.0.102.221(10.0.102.221:): This host has the latest relay log events.
Generating relay diff files from the latest slave succeeded.
10.0.102.221(10.0.102.221:): OK: Applying all logs succeeded. Slave started, replicating from 10.0.102.179(10.0.102.179:)
10.0.102.179(10.0.102.179:): Resetting slave info succeeded.
Master failover to 10.0.102.179(10.0.102.179:3306) completed successfully.
Sat Dec   ::  - [info] Sending mail..
Unknown option: conf

由mha切换日志可以看出，整个故障切换已经完成了。整个过程各个阶段核心切换逻辑简化后如下：

Phase 1：配置文件检查

Phase 2：非存活Master关闭服务

Phase 3：Master恢复

Phase 3.1：获取与Master延迟最小的Slave节点

Phase 3.2：生成Master与延迟最小的Slave节点的差异binlog并保存到manager节点

Phase 3.3：找出新的New Master，如果New Master不是最新的Slave节点，那么需要生成它们之间的差异Relay log

Phase 3.4: New Master恢复差异Relay log，随后获取Master binlog位点信息，然后恢复差异binlog日志

Phase 4: Slaves恢复

Phase 4.1：多线程生成延迟最小的Slave节点与其他一个或多个Slave差异Relay log

Phase 4.2：多线程恢复Slaves节点差异Relay log，然后change master到NEW MASTER节点

Phase 5：New Master清理Slave信息，并删除掉MHA配置文件中的选主信息防止误操作