MMM_CLUSTER部署文档

1   简介

CLUSTER允许一台服务器同时运行多个Agent客户端,或者运行多个Monitor

每个Agent都会有自己的端口,lock文件,日志文件

一个Agent只能对应一个Monitor,如果两个Monitor同时控制一个Agent,则会导致该Agent出现arp冲突,具体表现为:被设置的IP地址不稳定

同样,每个Monitor,也有自己的服务端口,lock文件,日志文件,以及自己控制的Agent

 

2   Agent CLUSTER

注意:tar包安装与yum安装后的配置文件略有不同

主要包含区别在于/etc/init.d/mysql-mmm-agent_CLUSTER文件,代码如下,大概意思是

1,指定一个集群节点的名字

2,指定lock文件

3,判断,如果CLUSTER不为空,则把@$CLUSTER传给/usr/sbin/mmm_agentd脚本,用以启动相应的agent cluster

 

# Cluster name (it can be empty for default cases)

CLUSTER='test1'

LOCKFILE='/var/lock/subsys/mysql-mmm-agent_test1'

prog='MMM Agent Daemon'

#-----------------------------------------------------------------------

# Paths

if [ "$CLUSTER" != "" ]; then

    MMMD_AGENT_BIN="/usr/sbin/mmm_agentd @$CLUSTER"

    MMMD_AGENT_PIDFILE="/var/run/mysql-mmm/mmm_agentd_$CLUSTER.pid"

else

    MMMD_AGENT_BIN="/usr/sbin/mmm_agentd"

    MMMD_AGENT_PIDFILE="/var/run/mysql-mmm/mmm_agentd.pid"

fi

 

#-----------------------------------------------------------------------

 

启动文件中设置好CLUSTER后,还需要修改两个配置文件

1, mmm_common_CLUSTER.conf,agent_port代表agent的启动端口,mysql_port代表控制的MySQL端口,pid_path的路径,名字用CLUSTER来区分.

 

active_master_role      writer

    cluster_interface       eth0

    agent_port              9911

    mysql_port              3306

    pid_path                /var/run/mysql-mmm/mmm_agentd_.pid

    bin_path                /usr/libexec/mysql-mmm/

    replication_user        replication

    replication_password    segD{cX56

    agent_user              mmm_agent

    agent_password          segD{cX56

 

2, mmm_agent_CLUSTER.conf,注意修改CLUSTER名字,对应到mmm_common_CLUSTER.conf文件

include mmm_common_CLUSTER.conf

this db1

 

 

3   Monitor CLUSTER

主要有三个文件,分别是

1,mmm_common_CLUSTER.conf,这个与Agent的一致

2,mmm_mon_CLUSTER.conf,监控端的配置文件

 

include mmm_common_test1.conf

 

    ip                  127.0.0.1

    port                9991

    pid_path            /var/run/mysql-mmm/mmm_mond_test1.pid

    bin_path            /usr/libexec/mysql-mmm

    status_path         /var/lib/mysql-mmm/mmm_mond_test1.status

    ping_interval       1

    ping_ips            192.168.250.2, 192.168.250.251, 192.168.250.252

    flap_duration       3600

    flap_count          3

    auto_set_online     0

    careful_startup     0

    mode                active

    wait_for_other_master  120

    # The kill_host_bin does not exist by default, though the monitor will

    # throw a warning about it missing.  See the section 5.10 "Kill Host

    # Functionality" in the PDF documentation.

    #

    # kill_host_bin     /usr/libexec/mysql-mmm/monitor/kill_host

    #

 

    monitor_user        mmm_monitor

    monitor_password    segD{cX56

 

    check_period  5

    trap_period   10

    timeout       2

    restart_after 10000

    max_backlog   600

 

debug 0

 

3,日志文件

mmm_mon_log_CLUSTER.conf

 

#log4perl.logger = ERROR, MMMLog, MailFatal

#log4perl.logger = FATAL, MMMLog

log4perl.logger = INFO, MMMLog, MailFatal

 

log4perl.appender.MMMLog = Log::Log4perl::Appender::File

log4perl.appender.MMMLog.Threshold = INFO

log4perl.appender.MMMLog.filename = /var/log/mysql-mmm/mmm_mond_CLUSTER.log

log4perl.appender.MMMLog.recreate = 1

log4perl.appender.MMMLog.layout = PatternLayout

log4perl.appender.MMMLog.layout.ConversionPattern = %d %5p %m%n

 

log4perl.appender.MailFatal = Log::Dispatch::Email::MailSender

log4perl.appender.MailFatal.Threshold = ERROR

log4perl.appender.MailFatal.from = [e-mail]

log4perl.appender.MailFatal.to = [e-mail]

log4perl.appender.MailFatal.buffered = 0

log4perl.appender.MailFatal.subject = FATAL error in mysql-mmm-monitor

log4perl.appender.MailFatal.layout = PatternLayout

log4perl.appender.MailFatal.layout.ConversionPattern = %d %m%n

 

4   实际部署 4.1     信息整理

CLUSTER	Function	HOSTNAME	IP	MySQL Port	Server ID	VIP Writer	VIP READER	note
test1	Agent	master1	192.168.250.251	3307	1	192.168.250.111
	Agent	master2	192.168.250.252	3307	2		192.168.250.112
test2	Agent	master1	192.168.250.251	3308	1	192.168.250.113
	Agent	master2	192.168.250.252	3308	2		192.168.250.114
test1/test2	Monitor	monitor253	192.168.250.253	-	-	-	-	-

 

4.2     MySQL帐号添加

GRANT REPLICATION SLAVE ON *.* TO 'replication'@'192.168.250.251' IDENTIFIED BY 'segD{cX56';

GRANT REPLICATION SLAVE ON *.* TO 'replication'@'192.168.250.252' IDENTIFIED BY 'segD{cX56';

GRANT SUPER, REPLICATION CLIENT, PROCESS ON *.* TO 'mmm_agent'@'192.168.250.251' IDENTIFIED BY 'segD{cX56';

GRANT SUPER, REPLICATION CLIENT, PROCESS ON *.* TO 'mmm_agent'@'192.168.250.252' IDENTIFIED BY 'segD{cX56';

GRANT REPLICATION CLIENT ON *.* TO 'mmm_monitor'@'192.168.250.253' IDENTIFIED BY 'segD{cX56';

 

4.3     MySQL主从配置

#重置binlog的命令RESET MASTER;

在3307端口的mysql执行,

 

CHANGE MASTER TO master_host='192.168.250.252', master_port=3307, master_user='replication', master_password='segD{cX56', master_log_file='mysql-bin.000001', master_log_pos=0;

START SLAVE;

 

CHANGE MASTER TO master_host='192.168.250.251', master_port=3307, master_user='replication', master_password='segD{cX56', master_log_file='mysql-bin.000001', master_log_pos=0;

START SLAVE;

 

在3308端口的mysql执行

CHANGE MASTER TO master_host='192.168.250.252', master_port=3308, master_user='replication', master_password='segD{cX56', master_log_file='mysql-bin.000001', master_log_pos=0;

START SLAVE;

 

CHANGE MASTER TO master_host='192.168.250.251', master_port=3308, master_user='replication', master_password='segD{cX56', master_log_file='mysql-bin.000001', master_log_pos=0;

START SLAVE;

 

4.4     Agent配置

以master1的3307端口为例,在master2上,3307端口的mysql,注意修改this db2

 

1.   下载yum源

wget

rpm -ivh epel-release-5-4.noarch.rpm

2.   安装mysql-mmm-agent

会自动安装依赖包

yum install mysql-mmm-agent

3.   修改配置文件(分别在master1和master2上修改)

mmm_common_test1.conf

 

active_master_role      writer

 

    cluster_interface       eth0

    agent_port              9991

    mysql_port              3307

    pid_path                /var/run/mysql-mmm/mmm_agentd_test1.pid

    bin_path                /usr/libexec/mysql-mmm/

    replication_user        replication

    replication_password    segD{cX56

    agent_user              mmm_agent

    agent_password          segD{cX56

 

    ip      192.168.250.251

    mode    master

    peer    db2

 

    ip      192.168.250.252

    mode    master

    peer    db1

 

    hosts   db1, db2

    ips     192.168.250.111

    mode    exclusive

 

    hosts   db1, db2

    ips     192.168.250.112

    mode    balanced

 

 

 

mmm_agent_test1.conf

 

include mmm_common_test1.conf

this db1

 

 

复制一个启动文件,并添加执行权限

注意,如果用restart重启,需要在stop和start之间添加sleep,否则在进程未杀死时,就启动了,导致重启失败.

 

cp /etc/init.d/mysql-mmm-agent /etc/init.d/mysql-mmm-agent-test1

chmod +x /etc/init.d/mysql-mmm-agent-test1

修改

vim /etc/init.d/mysql-mmm-agent-test1

 

# Cluster name (it can be empty for default cases)

CLUSTER='test1'

LOCKFILE='/var/lock/subsys/mysql-mmm-agent_test1'

prog='MMM Agent Daemon'

 

#-----------------------------------------------------------------------

# Paths

if [ "$CLUSTER" != "" ]; then

    MMMD_AGENT_BIN="/usr/sbin/mmm_agentd @$CLUSTER"

    MMMD_AGENT_PIDFILE="/var/run/mysql-mmm/mmm_agentd_$CLUSTER.pid"

else

    MMMD_AGENT_BIN="/usr/sbin/mmm_agentd"

    MMMD_AGENT_PIDFILE="/var/run/mysql-mmm/mmm_agentd.pid"

fi

#-----------------------------------------------------------------------

 

 

4.5     Monitor配置

主要有以下配置文件

mmm_common_test1.conf(与Agent一致)

 

mmm_mon_log_test1.conf

说明:MailFatal为邮件报警模块

perl日志级别,由低到高:INFO,WARN,ERROR,FATAL

 

log4perl.logger = INFO, MMMLog, MailFatal

#log4perl.logger = FATAL, MMMLog

 

log4perl.appender.MMMLog = Log::Log4perl::Appender::File

log4perl.appender.MMMLog.Threshold = INFO

log4perl.appender.MMMLog.filename = /var/log/mysql-mmm/mmm_mond_test1.log

log4perl.appender.MMMLog.recreate = 1

log4perl.appender.MMMLog.layout = PatternLayout

log4perl.appender.MMMLog.layout.ConversionPattern = %d %5p %m%n

 

log4perl.appender.MailFatal = Log::Dispatch::Email::MailSender

log4perl.appender.MailFatal.Threshold = ERROR

log4perl.appender.MailFatal.from = [e-mail]

log4perl.appender.MailFatal.to = [e-mail]

log4perl.appender.MailFatal.buffered = 0

log4perl.appender.MailFatal.subject = FATAL error in mysql-mmm-monitor from monitor253_test1

log4perl.appender.MailFatal.layout = PatternLayout

log4perl.appender.MailFatal.layout.ConversionPattern = %d %m%n

 

 

mmm_mon_test1.conf

 

include mmm_common_test1.conf

 

    ip                  127.0.0.1

    port                9981

    pid_path            /var/run/mysql-mmm/mmm_mond_test1.pid

    bin_path            /usr/libexec/mysql-mmm

    status_path         /var/lib/mysql-mmm/mmm_mond_test1.status

    ping_interval       1

    ping_ips            192.168.250.2, 192.168.250.251, 192.168.250.252

    flap_duration       3600

    flap_count          3

    auto_set_online     0

    careful_startup     0

    mode                active

    wait_for_other_master  120

 

    monitor_user        mmm_monitor

    monitor_password    segD{cX56

 

    check_period  5

    trap_period   10

    timeout       2

    restart_after 10000

    max_backlog   600

 

debug 0

 

 

启动文件/etc/init.d/mysql-mmm-monitor-test1

cp /etc/init.d/mysql-mmm-monitor /etc/init.d/mysql-mmm-monitor-test1

 

修改部分如下:

# Cluster name (it can be empty for default cases)

CLUSTER='test1'

LOCKFILE='/var/lock/subsys/mysql-mmm-monitor_test1'

prog='MMM Monitor Daemon'

 

if [ "$CLUSTER" != "" ]; then

        MMMD_MON_BIN="/usr/sbin/mmm_mond @$CLUSTER"

        MMMD_MON_PIDFILE="/var/run/mysql-mmm/mmm_mond_$CLUSTER.pid"

else

        MMMD_MON_BIN="/usr/sbin/mmm_mond"

        MMMD_MON_PIDFILE="/var/run/mysql-mmm/mmm_mond.pid"

fi

 

4.6     测试

在master1和master2上启动agent

/etc/init.d/mysql-mmm-agent-test1 start

 

在monitor253启动监控

/etc/init.d/mysql-mmm-monitor-test1 start

 

查看状态

[root@monitor253 mysql-mmm]# mmm_control @test1 show

  db1(192.168.250.251) master/AWAITING_RECOVERY. Roles:

  db2(192.168.250.252) master/ADMIN_OFFLINE. Roles:

 

[root@monitor253 mysql-mmm]# mmm_control @test1 set_online db1

OK: State of 'db1' changed to ONLINE. Now you can wait some time and check its new roles!

[root@monitor253 mysql-mmm]# mmm_control @test1 set_online db2

OK: State of 'db2' changed to ONLINE. Now you can wait some time and check its new roles!

[root@monitor253 mysql-mmm]# mmm_control @test1 show

  db1(192.168.250.251) master/ONLINE. Roles: writer(192.168.250.111)

  db2(192.168.250.252) master/ONLINE. Roles: reader(192.168.250.112)

 

[root@monitor253 mysql-mmm]# mmm_control @test1 checks

db2  ping         [last change: 2011/08/25 17:28:32]  OK

db2  mysql        [last change: 2011/08/25 17:36:19]  OK

db2  rep_threads  [last change: 2011/08/25 17:37:13]  OK

db2  rep_backlog  [last change: 2011/08/25 17:36:15]  OK: Backlog is null

db1  ping         [last change: 2011/08/25 17:28:32]  OK

db1  mysql        [last change: 2011/08/25 17:36:19]  OK

db1  rep_threads  [last change: 2011/08/25 17:37:10]  OK

db1  rep_backlog  [last change: 2011/08/25 17:36:15]  OK: Backlog is null

 

 

用vip可以连接

 

mysql -ummm_monitor -p'segD{cX56' -h192.168.250.111 -P3307

mysql -ummm_monitor -p'segD{cX56' -h192.168.250.112 -P3307