Orchestrator+Proxysql 实现自动导换+应用透明读写分离

前面一篇博客我介绍了如何通过Orachestrator+脚本实现自动导换。

这一章，我来演示一下如何通过Orchestrator+Proxysql 实现自动导换+应用透明读写分离

总体架构

可选架构1

实际生产系统所用架构应该类似于上图：

1）应用APP，通过VIP连接ProxySQL

2）ProxySQL为了实现高可用，可部署2个。通过VIP管理，即同一时候只有一个ProxySQL是活动并接受外界请求。

当proxysql出现故障时，通过keepyalive实现导换，业务切换到另一个Porxysql。

3）MySQL Master和Slave组成高可用MySQL集群

4）Orchestrator集群负责监控MySQL服务器的导换

可选架构2

如上架构与架构1的区别是，每个应用连接自有的Proxysql。如应用/proxysql crash，则相应的proxysql+app整体不可用。

演示步骤

简化架构

为了演示简单一些，在这里我简化了很多组件。结构图如下

测试机器

如上，为了测试我用了5台虚拟机

host01 192.168.56.103

host02 192.168.56.104

host03 192.168.56.105

arch  192.168.56.130

proxysql 192.168.56.131

ProxySQL安装

ProxySQL安装点 这里.

MySQL复制拓扑图

MySQL创建用户权限

testuser*是作为proxysql读写操作的。

monitor用户是proxysql用来监控MySQL活动状态的。

create user testuser_w@'%' identified by 'Testpass1.';
grant insert,update,delete on *.* to 'testuser_w'@'%';
create user testuser_r@'%' identified by 'Testpass1.';
grant select on *.* to 'testuser_r'@'%';
create user testuser_rw@'%' identified by 'Testpass1.';
grant select,insert,update,delete on *.* to 'testuser_rw'@'%';

create user 'monitor'@'%' identified by 'monitor';
grant all privileges on *.* to 'monitor'@'%';

proxysql启动并插入规则

[root@proxysql-server proxysql]# service proxysql.service start

Redirecting to /bin/systemctl start  proxysql.service.service

proxysql启动后，会开启两个端口。6032默认是用来管理，6033是默认用来连接mysql

[root@proxysql-server proxysql]# netstat -nlp|grep proxysql

tcp        0      0 0.0.0.0:6032            0.0.0.0:*               LISTEN      17840/proxysql

tcp        0      0 0.0.0.0:6033            0.0.0.0:*               LISTEN      17840/proxysql

登录proxysql管理端口配置数据

mysql -uadmin -padmin -h 127.0.0.1 -P 6032

插入如下配置。如下是指定600 hostgroup为只读，601端口为写入。（需要注意的是我这里的配置是master也做为读取结点）

其中scheduler的插入数据是做何用的，后面再讲述。

INSERT INTO mysql_servers (hostname,hostgroup_id,port,weight,max_replication_lag) VALUES ('192.168.56.103',601,3306,1000,60);
INSERT INTO mysql_servers (hostname,hostgroup_id,port,weight,max_replication_lag) VALUES ('192.168.56.104',601,3306,1000,60);
INSERT INTO mysql_servers (hostname,hostgroup_id,port,weight,max_replication_lag) VALUES ('192.168.56.105',601,3306,1000,60);
INSERT INTO mysql_servers (hostname,hostgroup_id,port,weight,max_replication_lag) VALUES ('192.168.56.103',600,3306,1000,0);
INSERT INTO mysql_replication_hostgroups(writer_hostgroup,reader_hostgroup) values(600,601);
LOAD MYSQL SERVERS TO RUNTIME; SAVE MYSQL SERVERS TO DISK;

insert into mysql_query_rules (username,destination_hostgroup,active) values('testuser_w',600,1);
insert into mysql_query_rules (username,destination_hostgroup,active) values('testuser_r',601,1);
insert into mysql_query_rules (username,destination_hostgroup,active,retries,match_digest) values('testuser_rw',601,1,3,'^SELECT');
LOAD MYSQL QUERY RULES TO RUNTIME;SAVE MYSQL QUERY RULES TO DISK;
insert into mysql_users (username,password,active,default_hostgroup,transaction_persistent) values ('testuser_w','Testpass1.',1,600,1);
insert into mysql_users (username,password,active,default_hostgroup,transaction_persistent) values ('testuser_r','Testpass1.',1,601,1);
insert into mysql_users (username,password,active,default_hostgroup,transaction_persistent) values ('testuser_rw','Testpass1.',1,600,1);
LOAD MYSQL USERS TO RUNTIME;SAVE MYSQL USERS TO DISK;

INSERT  INTO scheduler (id,interval_ms,filename,arg1) values (10,2000,"/var/lib/proxysql/server_monitor.pl","-u=admin -p=admin -h=127.0.0.1 -G=601 -P=6032 --debug=0  --log=/var/lib/proxysql/server_check");
LOAD SCHEDULER TO RUNTIME;SAVE SCHEDULER TO DISK;

插入成功后，查看服务器状态

如下可以看到，所以的服务器的状态都显示为ONLINE。

mysql> select * from mysql_servers;
+--------------+----------------+------+-----------+--------+--------+-------------+-----------------+---------------------+---------+----------------+---------+
| hostgroup_id | hostname       | port | gtid_port | status | weight | compression | max_connections | max_replication_lag | use_ssl | max_latency_ms | comment |
+--------------+----------------+------+-----------+--------+--------+-------------+-----------------+---------------------+---------+----------------+---------+
| 601          | 192.168.56.104 | 3306 | 0         | ONLINE | 1000   | 0           | 1000            | 60                  | 0       | 0              |         |
| 601          | 192.168.56.105 | 3306 | 0         | ONLINE | 1000   | 0           | 1000            | 60                  | 0       | 0              |         |
| 600          | 192.168.56.103 | 3306 | 0         | ONLINE | 1000   | 0           | 1000            | 60                  | 0       | 0              |         |
| 601          | 192.168.56.103 | 3306 | 0         | ONLINE | 1000   | 0           | 1000            | 60                  | 0       | 0              |         |
+--------------+----------------+------+-----------+--------+--------+-------------+-----------------+---------------------+---------+----------------+---------+

停止host01，查看导换后的变化

通过service mysqld stop，或是启动firewall，shutdwon mysql都可以。

host01无法连接后,mysql的拓扑结构变成了如下

查看proxysql状态数据的变化

如下，可以看到104的hostgroup变成了600（写入端口），host01（192.168.56.103）的状态变为了SHUNNED

mysql> select * from mysql_servers;
+--------------+----------------+------+-----------+--------+--------+-------------+-----------------+---------------------+---------+----------------+---------+
| hostgroup_id | hostname       | port | gtid_port | status | weight | compression | max_connections | max_replication_lag | use_ssl | max_latency_ms | comment |
+--------------+----------------+------+-----------+--------+--------+-------------+-----------------+---------------------+---------+----------------+---------+
| 601          | 192.168.56.103 | 3306 | 0         | ONLINE | 1000   | 0           | 1000            | 60                  | 0       | 0              |         |
| 601          | 192.168.56.105 | 3306 | 0         | ONLINE | 1000   | 0           | 1000            | 60                  | 0       | 0              |         |
| 600          | 192.168.56.104 | 3306 | 0         | ONLINE | 1000   | 0           | 1000            | 60                  | 0       | 0              |         |
| 601          | 192.168.56.104 | 3306 | 0         | ONLINE | 1000   | 0           | 1000            | 60                  | 0       | 0              |         |
+--------------+----------------+------+-----------+--------+--------+-------------+-----------------+---------------------+---------+----------------+---------+

mysql> select * from stats_mysql_connection_pool where hostgroup between 600 and 601 order by hostgroup,srv_host desc;
+-----------+----------------+----------+---------+----------+----------+--------+---------+-------------+---------+-------------------+-----------------+-----------------+------------+
| hostgroup | srv_host       | srv_port | status  | ConnUsed | ConnFree | ConnOK | ConnERR | MaxConnUsed | Queries | Queries_GTID_sync | Bytes_data_sent | Bytes_data_recv | Latency_us |
+-----------+----------------+----------+---------+----------+----------+--------+---------+-------------+---------+-------------------+-----------------+-----------------+------------+
| 600       | 192.168.56.104 | 3306     | ONLINE  | 0        | 1        | 1      | 242     | 1           | 18      | 0                 | 306             | 198             | 1282       |
| 601       | 192.168.56.105 | 3306     | ONLINE  | 0        | 1        | 1      | 0       | 1           | 127     | 0                 | 2159            | 1397            | 550        |
| 601       | 192.168.56.104 | 3306     | ONLINE  | 0        | 1        | 1      | 0       | 1           | 207     | 0                 | 3519            | 2277            | 1282       |
| 601       | 192.168.56.103 | 3306     | SHUNNED | 0        | 0        | 0      | 28      | 1           | 0       | 0                 | 0               | 0               | 389        |
+-----------+----------------+----------+---------+----------+----------+--------+---------+-------------+---------+-------------------+-----------------+-----------------+------------+
4 rows in set (0.01 sec)

通过mysql客户端，测试当前行为

如下，写一个简单的循环语句，测试mysql的行为。如下，可以发现写入操作都只会发送到新的host02机器。

只读操作，会发送到host02和host03机器。host01已经从读取列表内排除。

[root@proxysql-server proxysql]# while true; do mysql -u testuser_w -pTestpass1. -h 127.0.0.1 -P 6033 -e "select @@hostname"; sleep  1; done
mysql: [Warning] Using a password on the command line interface can be insecure.
+------------+
| @@hostname |
+------------+
| host02     |
+------------+
mysql: [Warning] Using a password on the command line interface can be insecure.
+------------+
| @@hostname |
+------------+
| host02     |
+------------+
^C
[root@proxysql-server proxysql]# while true; do mysql -u testuser_r -pTestpass1. -h 127.0.0.1 -P 6033 -e "select @@hostname"; sleep  1; done
mysql: [Warning] Using a password on the command line interface can be insecure.
+------------+
| @@hostname |
+------------+
| host03     |
+------------+
mysql: [Warning] Using a password on the command line interface can be insecure.
+------------+
| @@hostname |
+------------+
| host02     |
+------------+

当host01重新启动后，会发生什么？

细心的读者，可能会发生，mysql_servers列表内，并没有排除host01。那如果host01启动后，会发生什么呢？

结果就是只读操作会继续发送到host01机器。

因为实际上host01机器在导换后，并不属于复制集群了。

如果只读操作继续发送到host01，肯定会读到旧的数据。造成应用出现问题。

那我们如何避免出现此问题呢？方法就是通过上面的脚本来把host01排除出复制集群。

后续的修复需要DBA来干预，如加入复制集群，等复制同步完成后，再开放host01可以作为读结点。

增加一个定时任务

脚本做如下检查

• read_only=ON – mysql服务器是否为 read-only (Slave，read-only为ON)
• repl_lag  is NULL – Master，这个参数值应该为NULL (如果seconds_behind_master没有定义, ProxySQL 会报告repl_lag 为 NULL)

如果 read_only=ON, 这意味者，这台服务器当前不是master。但如果repl_lag=NULL, 这就意味得服务器没有开启复制，而且很有可能以前是master. 所以，就必须把它从Hostgroup内移除.

INSERT  INTO scheduler (id,interval_ms,filename,arg1) values (10,2000,"/var/lib/proxysql/server_monitor.pl","-u=admin -p=admin -h=127.0.0.1 -G=601 -P=6032 --debug=0  --log=/var/lib/proxysql/server_check");
LOAD SCHEDULER TO RUNTIME;SAVE SCHEDULER TO DISK;

检查mysql服务器状态

当host01重新启动后。

如下，可以看到host01（192.168.56.103）的状态是在线，但是hostgroup修改成为了9601（server_monitor.pl脚本实现就是通过修改   New hostgroup=9000+ old hostgroup）。如此一来，发往601只读hostgroup里，就排除了host01。

mysql> select * from mysql_servers;
+--------------+----------------+------+-----------+--------+--------+-------------+-----------------+---------------------+---------+----------------+---------+
| hostgroup_id | hostname       | port | gtid_port | status | weight | compression | max_connections | max_replication_lag | use_ssl | max_latency_ms | comment |
+--------------+----------------+------+-----------+--------+--------+-------------+-----------------+---------------------+---------+----------------+---------+
| 9601         | 192.168.56.103 | 3306 | 0         | ONLINE | 1000   | 0           | 1000            | 60                  | 0       | 0              |         |
| 600          | 192.168.56.104 | 3306 | 0         | ONLINE | 1000   | 0           | 1000            | 60                  | 0       | 0              |         |
| 601          | 192.168.56.105 | 3306 | 0         | ONLINE | 1000   | 0           | 1000            | 60                  | 0       | 0              |         |
| 601          | 192.168.56.104 | 3306 | 0         | ONLINE | 1000   | 0           | 1000            | 60                  | 0       | 0              |         |
+--------------+----------------+------+-----------+--------+--------+-------------+-----------------+---------------------+---------+----------------+---------+
4 rows in set (0.00 sec)

server_monitor.pl完整脚本

#!/usr/bin/perl
# This tool is "fat-packed": most of its dependent modules are embedded
# in this file. 

package remove_old_master;
use Time::HiRes qw(gettimeofday);
use strict;
use DBI;
use Getopt::Long;
use Pod::Usage;
use Data::Dumper;

$Getopt::Long::ignorecase = 0;
my $Param = {};
my $user = "admin";
my $pass = "admin";
my $help = '';
my $host = 'localhot' ;
my $debug = 0 ;
my %hostgroups;

my %processState;
my %processCommand;
my @HGIds;

######################################################################
#Local functions
######################################################################

sub URLDecode {
my $theURL = $_[0];
$theURL =~ tr/+/ /;
$theURL =~ s/%([a-fA-F0-9]{2,2})/chr(hex($1))/eg;
$theURL =~ s/<!--(.|\n)*-->//g;
return $theURL;
}
sub URLEncode {
my $theURL = $_[0];
$theURL =~ s/([\W])/"%" . uc(sprintf("%2.2x",ord($1)))/eg;
return $theURL;
}

# return a proxy object
sub get_proxy($$$$){
my $dns = shift;
my $user = shift;
my $pass = shift;
my $debug = shift;
my $proxynode = ProxySqlNode->new();
$proxynode->dns($dns);
$proxynode->user($user);
$proxynode->password($pass);
$proxynode->debug($debug);

return $proxynode;

}

sub main{
# ============================================================================
#+++++ INITIALIZATION
# ============================================================================

if($#ARGV < 3){
#given a ProxySQL scheduler
#limitation we will pass the whole set of params as one
# and will split after
@ARGV = split('\ ',$ARGV[0]);
}
$Param->{user} = '';
$Param->{log} = undef ;
$Param->{password} = '';
$Param->{host} = '';
$Param->{port} = '127.0.0.1';
$Param->{debug} = 0;
$Param->{processlist} = 0;
$Param->{OS} = $^O;
$Param->{main_segment} = 0;
$Param->{retry_up} = 0;
$Param->{retry_down} = 0;
$Param->{print_execution} = 0;
$Param->{development} = 0;
$Param->{hgid} = 0;

my $run_pid_dir = "/tmp" ;

#if (
GetOptions(
'user|u:s' => \$Param->{user},
'password|p:s' => \$Param->{password},
'host|h:s' => \$host,
'port|P:i' => \$Param->{port},
'debug|d:i' => \$Param->{debug},
'log:s' => \$Param->{log},
'main_segment|S:s'=> \$Param->{main_segment},
'retry_up:i' => \$Param->{retry_up},
'retry_down:i' => \$Param->{retry_down},
'execution_time:i' => \$Param->{print_execution},
'development:i' => \$Param->{development},
'active_failover' => \$Param->{active_failover},
'hgid|G:i' => \$Param->{hgid},

'help|?' => \$Param->{help}

) or pod2usage(2);
pod2usage(-verbose => 2) if $Param->{help};

die print Utils->print_log(1,"Option --hgid not specified.\n") unless defined($Param->{hgid});
die print Utils->print_log(1,"Option --host not specified.\n") unless defined $Param->{host};
die print Utils->print_log(1,"Option --user not specified.\n") unless defined $Param->{user};
die print Utils->print_log(1,"Option --port not specified.\n") unless defined $Param->{port};
die "Option --log not specified. We need a place to log what is going on, don't we?\n" unless defined $Param->{log};
print Utils->print_log(2,"Option --log not specified. We need a place to log what is going on, don't we?\n") unless defined $Param->{log};

if($Param->{debug}){
Utils::debugEnv();
}

$Param->{host} = URLDecode($host);
my $dsn = "DBI:mysql:host=$Param->{host};port=$Param->{port}";
if(defined $Param->{user}){
$user = "$Param->{user}";
}
if(defined $Param->{password}){
$pass = "$Param->{password}";
}
my $hg =$Param->{hgid};
$hg =~ s/[\:,\,]/_/g;
my $base_path = "${run_pid_dir}/proxysql_galera_check_${hg}.pid";

#============================================================================
# Execution
#============================================================================
if(defined $Param->{log}){
open(FH, '>>', $Param->{log}."_".$hg.".log") or die Utils->print_log(1,"cannot open file");
#select FH;
}

if($Param->{development} < 1){
if(!-e $base_path){
`echo "$$ : $hg" > $base_path`
}
else{
print Utils->print_log(1,"Another process is running using the same HostGroup and settings,\n Or orphan pid file. check in $base_path");
exit 1;
}
}

# for test only purpose comment for prod

#my $xx =1;
#my $y =0;
#$xx=20000000 if($Param->{development} > 0);

#while($y < $xx){
#++$y ;

my $start = gettimeofday();
if($Param->{debug} >= 1){
print Utils->print_log(3,"START EXECUTION\n");
}

my $proxy_sql_node = get_proxy($dsn, $user, $pass ,$Param->{debug}) ;

$proxy_sql_node->retry_down($Param->{retry_down});
$proxy_sql_node->move_node($proxy_sql_node,$Param->{hgid});

$proxy_sql_node->connect();

my $end = gettimeofday();
print Utils->print_log(3,"END EXECUTION Total Time:".($end - $start) * 1000 ."\n\n") if $Param->{print_execution} >0;

$proxy_sql_node->disconnect();

#debug braket
sleep 2 if($Param->{development} > 0);
}
if(defined $Param->{log}){
close FH; # in the end
}

`rm -f $base_path`;

exit(0);

}

# ############################################################################
# Run the program.
# ############################################################################
exit main(@ARGV);

{
package ProxySqlNode;
sub new {
my $class = shift;

my $SQL_get_monitor = "select variable_name name,variable_value value from global_variables where variable_name in( 'mysql-monitor_username','mysql-monitor_password','mysql-monitor_read_only_timeout' ) order by 1";
my $SQL_get_hostgroups = "select distinct hostgroup_id hg_isd from mysql_servers order by 1;";
my $SQL_get_rep_hg = "select writer_hostgroup,reader_hostgroup from mysql_replication_hostgroups order by 1;";

# Variable section for looping values
#Generalize object for now I have conceptualize as:
# Proxy (generic container)
# Proxy->{DNS} conenction reference
# Proxy->{PID} processes pid (angel and real)
# Proxy->{hostgroups}
# Proxy->{user} This is the user name
# Proxy->{password}
# Proxy->{port} node status (OPEN 0,Primary 1,Joiner 2,Joined 3,Synced 4,Donor 5)

my $self = {
_dns => undef,
_pid => undef,
_user => undef,
_password => undef,
_port => undef,
_hgid => undef,
_monitor_user => undef,
_monitor_password => undef,
_SQL_get_monitor => $SQL_get_monitor,
_SQL_get_hg=> $SQL_get_hostgroups,
_SQL_get_replication_hg=> $SQL_get_rep_hg,
_dbh_proxy => undef,
_check_timeout => 100, #timeout in ms
_retry_down => 0, # number of retry on a node before declaring it as failed.

};
bless $self, $class;
return $self;

}

sub retry_down{
my ( $self, $in ) = @_;
$self->{_retry_down} = $in if defined($in);
return $self->{_retry_down};
}

sub debug{
my ( $self, $debug ) = @_;
$self->{_debug} = $debug if defined($debug);
return $self->{_debug};
}

sub dns {
my ( $self, $dns ) = @_;
$self->{_dns} = $dns if defined($dns);
return $self->{_dns};
}

sub dbh_proxy{
my ( $self, $dbh_proxy ) = @_;
$self->{_dbh_proxy} = $dbh_proxy if defined($dbh_proxy);
return $self->{_dbh_proxy};
}

sub pid {
my ( $self, $pid ) = @_;
$self->{_pid} = $pid if defined($pid);
return $self->{_pid};
}

sub user{
my ( $self, $user ) = @_;
$self->{_user} = $user if defined($user);
return $self->{_user};
}

sub password {
my ( $self, $password ) = @_;
$self->{_password} = $password if defined($password);
return $self->{_password};
}

sub monitor_user{
my ( $self, $monitor_user ) = @_;
$self->{_monitor_user} = $monitor_user if defined($monitor_user);
return $self->{_monitor_user};
}

sub monitor_password {
my ( $self, $monitor_password ) = @_;
$self->{_monitor_password} = $monitor_password if defined($monitor_password);
return $self->{_monitor_password};
}

sub port {
my ( $self, $port ) = @_;
$self->{_port} = $port if defined($port);
return $self->{_port};
}

sub hgid {
my ( $self, $hgid ) = @_;
$self->{_hgid} = $hgid if defined($hgid);
print Dumper($hgid);
return $self->{_hgid};
}

sub check_timeout{
my ( $self, $check_timeout ) = @_;
$self->{_check_timeout} = $check_timeout if defined($check_timeout);
return $self->{_check_timeout};
}

sub move_node{
my ( $self,$proxynode,$hgid ) = @_;

my $dbh = Utils::get_connection($self->{_dns}, $self->{_user}, $self->{_password},' ');
$self->{_dbh_proxy} = $dbh;

my $SQL_get_old_master= "SELECT ro.hostname,
 ro.port,
 ms.hostgroup_id
 FROM (select *, max(time_start_us) from monitor.mysql_server_read_only_log group by hostname,port) as ro
 LEFT JOIN (select *,max(time_start_us) from monitor.mysql_server_replication_lag_log group by hostname,port ) as lag ON ro.hostname=lag.hostname
 AND ro.port=lag.port
 LEFT JOIN mysql_servers ms ON ro.hostname=ms.hostname
 AND ro.port=ms.port,
 mysql_replication_hostgroups hg
 WHERE read_only=1
 AND repl_lag IS NULL
 AND ms.hostgroup_id='$hgid'
 AND ms.hostgroup_id = hg.reader_hostgroup
 AND ms.status='ONLINE'
 GROUP BY ro.hostname,
 ro.port;";

my $sth = $dbh->prepare($SQL_get_old_master);
my $ref;
$sth->execute();
while ($ref = $sth->fetchrow_arrayref()) {
print "@$ref[0] @$ref[1]\n";
if (@$ref[0] ne '') {
move_node_down_hg_change($proxynode,@$ref[0],@$ref[1],$Param->{hgid});
}

}
}

#Connect method connect an populate the cluster returns the Galera cluster
sub connect{
my ( $self, $port ) = @_;
my $dbh = Utils::get_connection($self->{_dns}, $self->{_user}, $self->{_password},' ');
$self->{_dbh_proxy} = $dbh;

# get monitor user/pw
my $cmd = $self->{_SQL_get_monitor};

my $sth = $dbh->prepare($cmd);
$sth->execute();
while (my $ref = $sth->fetchrow_hashref()) {
if($ref->{'name'} eq 'mysql-monitor_password' ){$self->{_monitor_password} = $ref->{'value'};}
if($ref->{'name'} eq 'mysql-monitor_username' ) {$self->{_monitor_user} = $ref->{'value'};}
if($ref->{'name'} eq 'mysql-monitor_read_only_timeout' ) {$self->{_check_timeout} = $ref->{'value'};}

}
if($self->debug >=1){print Utils->print_log(3," Connecting to ProxySQL " . $self->{_dns}. "\n" ); }

}
sub disconnect{
my ( $self, $port ) = @_;
$self->{_dbh_proxy}->disconnect;

}

#move a node to a maintenance HG ((9000 + HG id))
sub move_node_down_hg_change{
my ($proxynode, $host,$port,$hgid) = @_;

if($hgid > 9000) {return 1;}

my $node_sql_command = "SET GLOBAL READ_ONLY=1;";
my $proxy_sql_command =" UPDATE mysql_servers SET hostgroup_id=".(9000 + $hgid)." WHERE hostgroup_id=$hgid AND hostname='$host' AND port='$port'";
print "$proxy_sql_command\n";
$proxynode->{_dbh_proxy}->do($proxy_sql_command) or die "Couldn't execute statement: " . $proxynode->{_dbh_proxy}->errstr;
$proxynode->{_dbh_proxy}->do("LOAD MYSQL SERVERS TO RUNTIME") or die "Couldn't execute statement: " . $proxynode->{_dbh_proxy}->errstr;
print Utils->print_log(2," Move node:"
." SQL:" .$proxy_sql_command
."\n" );

}

}

{
package ProxySqlHG;
sub new {
my $class = shift;

my $self = {
_id => undef, #
_type => undef, # available types: w writer; r reader ; mw maintance writer; mr maintenance reader
_size => 0,
};
bless $self, $class;
return $self;
}

sub id {
my ( $self, $id ) = @_;
$self->{_id} = $id if defined($id);
return $self->{_id};
}

sub type {
my ( $self, $type ) = @_;
$self->{_type} = $type if defined($type);
return $self->{_type};
}
sub size {
my ( $self, $size ) = @_;
$self->{_size} = $size if defined($size);
return $self->{_size};
}

}

{
package Utils;
use Time::HiRes qw(gettimeofday);
#============================================================================
## get_connection -- return a valid database connection handle (or die)
## $dsn -- a perl DSN, e.g. "DBI:mysql:host=ltsdbwm1;port=3311"
## $user -- a valid username, e.g. "check"
## $pass -- a matching password, e.g. "g33k!"

sub get_connection($$$$) {
my $dsn = shift;
my $user = shift;
my $pass = shift;
my $SPACER = shift;
my $dbh = DBI->connect($dsn, $user, $pass);

if (!defined($dbh)) {
print Utils->get_current_time ."[ERROR] Cannot connect to $dsn as $user\n";
# die();
return undef;
}

return $dbh;
}

#Prrint time from invocation with milliseconds
sub get_current_time{
use POSIX qw(strftime);
my $t = gettimeofday();
my $date = strftime "%Y/%m/%d %H:%M:%S", localtime $t;
$date .= sprintf ".%03d", ($t-int($t))*1000; # without rounding

return $date;
}

#prit all environmnt variables
sub debugEnv{
my $key = keys %ENV;
foreach $key (sort(keys %ENV)) {

print $key, '=', $ENV{$key}, "\n";

}

}

#Print a log entry
sub print_log($$){
my $log_level = $_[1];
my $text = $_[2];
my $log_text = "[ - ] ";

SWITCH: {
if ($log_level == 1) { $log_text= "[ERROR] "; last SWITCH; }
if ($log_level == 2) { $log_text= "[WARN] "; last SWITCH; }
if ($log_level == 3) { $log_text= "[INFO] "; last SWITCH; }
if ($log_level == 4) { $log_text= "[DEBUG] "; last SWITCH; }
}
return Utils::get_current_time.":".$log_text.$text;

}

#trim a string
sub trim {
my $s = shift;
$s =~ s/^\s+|\s+$//g;
return $s
};

}

# ############################################################################
# Documentation
# #################
=pod

=head1 NAME
server_monitor.pl

=head1 OPTIONS
    server_monitor.pl -u=admin -p=admin -h=192.168.1.50 -G=601 -P=3310 --debug=0 --log <full_path_to_file> --help
    sample [options] [file ...]
    Options:
      -u|user user to connect to the proxy
      -p|password Password for the proxy
      -h|host Proxy host
      -G Hostgroup ID with role definition.
      --retry_down The number of loop/test the check has to do before moving a node Down (default 0)
      --debug
      --log Full path to the log file ie (/var/log/proxysql/galera_check_) the check will add the identifier for the specific HG.
      -help help message

=head1 DESCRIPTION

Server monitor is a script to remove the server from the HostGroup which are not part of the replicaset anymore.
The script monitors the read_only flag and the repl_lag. If read_only=1 and repl_lag is NULL it removes the server from the HostGroup.

Why we need this?

Example: If we promote a new master and the old master won't be the a slave of the ne master, the ProxySQL will thought that server now is part of the
read hostgroup and going to send reads. But the old master does not part of the replicaset anymore. ProxySQL should not send any traffic.

=head1 Configure in ProxySQL

INSERT INTO scheduler (id,interval_ms,filename,arg1) values (10,2000,"/var/lib/proxysql/server_monitor.pl","-u=admin -p=admin -h=192.168.1.50 -G=601 -P=3310 --retry_down=2 --debug=0 --log=/var/lib/proxysql/server_check");
LOAD SCHEDULER TO RUNTIME;SAVE SCHEDULER TO DISK;

update scheduler set arg1="-u=admin -p=admin -h=192.168.1.50 -G=601 -P=3310 --debug=1 --log=/var/lib/proxysql/server_check" where id =10;
LOAD SCHEDULER TO RUNTIME;SAVE SCHEDULER TO DISK;

delete from scheduler where id=10;
LOAD SCHEDULER TO RUNTIME;SAVE SCHEDULER TO DISK;
=cut

注意事项

因为proxysql通过read_only变量来进行消息的转发控制，所以我们启动服务器的时候，一定要以read_only方式进行启动(即便是master)。如此就能避免很多类似多点同时写入的问题。

参考资料

https://www.percona.com/blog/2016/11/09/orchestrator-and-proxysql/