[MariaDB]MHA高可用部署-實驗
阿新 • • 發佈：2020-01-09
目錄
- 一、簡介
- 1.1MHA角色
- 二、MHA的工具
- 三、MHA部署過程
- 3.1.1 配置
- 3.1.2 環境規劃
- 3.1.3 配置一主多從
- 3.2 MHA配置
- 3.2.1 master許可權授予
- 3.2.2 ssh互信
- 3.2.3 安裝mha包
- 3.2.4 MHA管理節點配置
- 3.2.5 MHA節點檢測
- 3.2.6 MHA啟動
- 3.2.7 MHA模擬故障
- 3.2.8 修復已損壞的節點加入MHA中
- MHA 問題集錦
一、簡介
MHA的邏輯是，為了保證其MySQL的高可用，會有一個StandBy狀態的master.在mysql故障切換的過程中，MHA 能做到在 0~30 秒內自動完成資料庫的故障切換操作，並且在進行故障切換的過程中，MHA 能最大程度的保證資料的一致性，以達到相對意義上的高可用。
1.1MHA角色
如下圖，整個 MHA 架構分為
- MHA Manager 節點
- MHA Node 節點
其中 MHA Manager 節點是單點部署，MHA Node 節點是部署在每個需要監控的 MySQL 叢集節點上的。MHA Manager 會定時探測叢集中的 Master 節點，當 Master 出現故障時，它可以自動將最新資料的 Standby Master 或 Slave 提升為新的 Master，然後將其他的 Slave 重新指向新的 Master。
二、MHA的工具
Manager節點：
masterha_check_ssh：MHA 依賴的 ssh 環境監測工具；masterha_check_repl：MYSQL 複製環境檢測工具；masterga_manager：MHA 服務主程式；masterha_check_status：MHA 執行狀態探測工具；masterha_master_monitor：MYSQL master 節點可用性監測工具；masterha_master_swith:master：節點切換工具；masterha_conf_host：新增或刪除配置的節點；masterha_stop：關閉 MHA 服務的工具。
Node節點：（這些工具通常由MHA Manager的指令碼觸發，無需人為操作）
save_binary_logs：儲存和複製 master 的二進位制日誌；apply_diff_relay_logs：識別差異的中繼日誌事件並應用於其他 slave；purge_relay_logs：清除中繼日誌（不會阻塞 SQL 執行緒）；
自定義擴充套件：
secondary_check_script：通過多條網路路由檢測master的可用性；master_ip_failover_script：更新application使用的masterip；report_script：傳送報告;init_conf_load_script：載入初始配置引數；master_ip_online_change_script;更新master節點ip地址。
三、MHA部署過程
3.1.1 配置
MHA 對 MYSQL 複製環境有特殊要求，例如各節點都要開啟二進位制日誌及中繼日誌，各從節點必須顯示啟用其read-only屬性，並關閉relay_log_purge功能等，這裡對配置做事先說明。
3.1.2 環境規劃
| 機器名 | IP | 角色 | 備註 |
|---|---|---|---|
| manager | 172.30.200.100 | manager控制器 | 用於管理和故障切換 |
| master | 172.30.200.101 | 資料庫主伺服器 | 開啟binlog,relay-log。關閉relay_log_purge |
| slave1 | 172.30.200.102 | 資料庫從伺服器 | 開啟binlog,relay-log。關閉relay_log_purge |
| slave2 | 172.30.200.103 | 資料庫從伺服器 | 開啟binlog,relay-log。關閉relay_log_purge |
在各個節點的/etc/hosts檔案配置內容新增如下:
[root@localhost ~]# cat /etc/hosts
127.0.0.1 localhost localhost.localdomain localhost4 localhost4.localdomain4
::1 localhost localhost.localdomain localhost6 localhost6.localdomain6
172.30.200.100 arpmgr
172.30.200.101 arpmaster
172.30.200.102 arpslave1
172.30.200.103 arpslave2建立binlog的目錄
mkdir -p /data/mysqldata/binlog
chown -R mysql:mysql /data/mysqldata/binlog101節點配置:
server-id = 200101
log-bin = /data/mysqldata/binlog/mysql-bin
binlog_format= row
max_binlog_size= 512m
relay-log = /data/mysqldata/binlog/relay-bin
expire-logs-days = 14
lower_case_table_names = 1
character-set-server = utf8
log_slave_updates = 1102節點配置:
server-id = 200102
log-bin = /data/mysqldata/binlog/mysql-bin
binlog_format= row
max_binlog_size= 512m
relay-log = /data/mysqldata/binlog/relay-bin
expire-logs-days = 14
read_only = ON
relay_log_purge = 0
lower_case_table_names = 1
character-set-server = utf8
log_slave_updates = 1103節點配置:
server-id = 200103
log-bin = /data/mysqldata/binlog/mysql-bin
binlog_format= row
max_binlog_size= 512m
relay-log = /data/mysqldata/binlog/relay-bin
read_only = ON
relay_log_purge = 0
expire-logs-days = 14
lower_case_table_names = 1
character-set-server = utf8
log_slave_updates = 13.1.3 配置一主多從
master節點配置:
MariaDB [(none)]>grant replication slave,replication client on *.* to 'repl'@'172.30.200.%' identified by 'repl7101';
MariaDB [(none)]> show master status;
+------------------+----------+--------------+------------------+
| File | Position | Binlog_Do_DB | Binlog_Ignore_DB |
+------------------+----------+--------------+------------------+
| mysql-bin.000001 | 548 | | |
+------------------+----------+--------------+------------------+
1 row in set (0.000 sec)slave節點配置:
grant replication slave,replication client on *.* to 'repl'@'172.30.200.%' identified by 'repl7101';
change master to master_host='172.30.200.101',
master_user='repl',
master_password='repl7101',
master_log_file='mysql-bin.000001',
master_log_pos=548;
start slave;
show slave status\G;至此，一主多從配置完畢。
3.2 MHA配置
3.2.1 master許可權授予
可以在所有節點上面配置，其擁有管理許可權，目前只需在master結點上設定許可權:
grant all on *.* to 'mhaadmin'@'172.30.%.%' identified by 'mha7101';
grant all on *.* to 'mhaadmin'@'arpmgr' identified by 'mha7101';3.2.2 ssh互信
四個節點都執行如下語句:
ssh-keygen -t rsa
ssh-copy-id -i .ssh/id_rsa.pub root@arpmgr然後在arpmgr結點上面，可以看到authorized_keys檔案的資訊內容如下:
[root@localhost .ssh]# cat authorized_keys
ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQDY3yFhR5uzcFEpd+q+1Uw/cRF9ZRraygms7OZoefLFzY/ydSi6yYuCighG8WquvRep7XDNjFI71HAUagSoXiyPoCe1lqEnzpxSc+fQpIeQqEhUmLJ2bk+R83EskzwRGh+S/D4yp/swWz1vRgUGoTWevLCs33q7ZrsM8i+jB0uwZmzOV+CyQAPW9vLkRjZa4y1sx65lbR0HbdTQWQYZ4IyZauoU8XQjAIOs/CdLw2nBt8dPO53jT7NS7Ywx6eu/Wj9k/sYVVZT3jTb+pBIVs+Du5+tdUDX5aLKzxINpLlqNhorNevoC9iE0Ame1qvYonQfyWQ52Ae0y+58vFfG6PyV3 [email protected]
ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQC/ZPihYSC6ArawKRU75aQRVSFsQ5S89SrYHGWdzyluB4spj+UDUmWH1kLGYr715/HD5hh22KdLmIs7R4jviOeao1HK52fpMvklYaNtYRHuV63Zkg5sOLvLfhrHdta9wuHlW1NyWx75+wIl2LvKBRtnSddwf5ZvitJ/kChf2gpNhHAWidyjGsPoJdr0OBCNHvz1y6oON6cnMb07ExaIjptRnkbCOU0QSVjFq4+Jmh8zTTbJC2up50s15gSfWXH0+WLXmJXJGkvgHdSYqw4vJt/l25f5qAKKZsfnyfC0iyct4GyHPF6trpvQ/c2lqr/Rg4xLWgdxlyt4aBJYl5adIRK/ [email protected]
ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQDba26wV0KwQNTb4pKuiFDCcVMNRLGMXSiJC8ucN4/KIqzoOYJ747QL8GL5F8ePnRaZ1rtOwdjnlTiC0a4Tcg4JLs+JSnJgzvepuixmGgSJfLbJ36iN1WFh6fP2GZEDdR7Qum4sBUpQyYJ20Kf9rKfQQv2wq6csK5IlFk/OoO+zTySauLnYvRxvKY2avVDXPPFJvpqimKXn59MIAoJr6YEKvncbYyqvrSgUy7klZDys9IIjYcWfO7VKjQ5bwbHrrKtNbedME+KPQld7e8ZVL66Omik4Z6ip7DQEHRKWMmuBIpL99AgOOjPLbzJFWLUPOwvy3DtmEBnZ+0NVf/1obC11 [email protected]
ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQCwrgZtGC31EgixeY4SVl4h64m1r8LdL3hM4Be2/I+6Xw7hCzZyKKTAFgz9W/ukfx6WmZwoqp1VO/7Jp6KO1FhYOi5u0q6J1KIObFNp+3E6cB2P0q39WqmZpQ9cNPYrbs9U2Ej0L0JwUtf/xLh334PaSlv/LcNy+p1dWya2OqsBeraiXZ4MgEBzcb+0twkpfpD327VgT/mRHPmA6fPRJOOJti1u4isHeotE4i13YIqQYfBfmbfiLdXKAvgI8FuTf0i91Re/FUBOgBfBcJbqIQNR0Nh5wZ/LvNxkstDQvypZIZwiK+wN+aZZOQ7jF/+997Z9QQleC9OOoHOJR7+fisLb [email protected]正好有四個形如ssh-rsa相關 的金鑰資訊。
把如上的公鑰資訊，拷貝到其餘四臺伺服器上面:
scp authorized_keys root@arpmaster:~/.ssh/
scp authorized_keys root@arpslave1:~/.ssh/
scp authorized_keys root@arpslave2:~/.ssh/測試ssh是否可用
[root@localhost .ssh]# ssh arpmaster
[root@localhost ~]# ssh arpslave1
[root@localhost ~]# ssh arpslave2
[root@localhost ~]# ssh arpmgr3.2.3 安裝mha包
mha安裝包分為兩個，一個是node,另外一個是manager。
四個節點安裝:mha4mysql-node-0.57-0.el7.centos.noarch.rpm
管理節點安裝:mha4mysql-manager-0.57-0.el7.centos.noarch.rpm
在安裝`mha4mysql-node-0.57-0.el7.centos.noarch.rpm過程中,有對perl-DBD-mysql,perl-DBI前置依賴,安裝步驟如下:
yum install perl-DBD-mysql perl-DBI在安裝`mha4mysql-manager-0.57-0.el7.centos.noarch.rpm過程中,有對perl前置依賴,安裝步驟如下:
安裝yum 擴充套件包
yum install https://dl.fedoraproject.org/pub/epel/epel-release-latest-7.noarch.rpm
yum install perl-ExtUtils-CBuilder perl-ExtUtils-MakeMaker perl-Config-Tiny perl-Log-Dispatch-* perl-Parallel-ForkManager然後安裝資訊，都成功，如下:
[root@localhost ~]# rpm -ivh mha4mysql-node-0.57-0.el7.centos.noarch.rpm
準備中... ################################# [100%]
正在升級/安裝...
1:mha4mysql-node-0.58-0.el7.centos ################################# [100%]
[root@localhost ~]# rpm -ivh mha4mysql-manager-0.57-0.el7.centos.noarch.rpm
準備中... ################################# [100%]
正在升級/安裝...
1:mha4mysql-manager-0.58-0.el7.cent################################# [100%]0.58中有一個super_read_only不可用在mariadb，所以使用0.57版本。
3.2.4 MHA管理節點配置
[root@localhost ~]# cd /etc/mha_master/
[root@localhost mha_master]# vi /etc/mha_master/mha.cnf配置檔案內容如下:
[server default]
user=mhaadmin
password=mha7101
manager_workdir=/etc/mha_master/app1
manager_log=/etc/mha_master/manager.log
remote_workdir=/data/mha_master/app1
repl_user=repl
repl_password=repl7101
ping_interval=1
[server1]
hostname=172.30.200.101
ssh_port=22
[server2]
hostname=172.30.200.102
ssh_port=22
candidate_master=1
[server3]
hostname=172.30.200.103
ssh_port=22
no_master=13.2.5 MHA節點檢測
在管理節點檢測ssh連通性如下:
[root@localhost ~]# masterha_check_ssh -conf=/etc/mha_master/mha.cnf有如下日誌,代表正常:
Thu Jan 9 14:43:09 2020 - [warning] Global configuration file /etc/masterha_default.cnf not found. Skipping. Thu Jan 9 14:43:09 2020 - [info] Reading application default configuration from /etc/mha_master/mha.cnf.. [email protected](172.30.200.103:22) to [email protected](172.30.200.101:22).. Thu Jan 9 14:43:11 2020 - [debug] ok. Thu Jan 9 14:43:11 2020 - [debug] Connecting via SSH from [email protected](172.30.200.103:22) to [email protected](172.30.200.102:22).. Thu Jan 9 14:43:11 2020 - [debug] ok. Thu Jan 9 14:43:12 2020 - [info] All SSH connection tests passed successfully.檢測MySQL replication是否正常
masterha_check_repl --conf=/etc/mha_master/mha.cnf有如下日誌，說明正常:
Thu Jan 9 14:44:54 2020 - [info] Slaves settings check done. Thu Jan 9 14:44:54 2020 - [info] 172.30.200.101(172.30.200.101:3306) (current master) +--172.30.200.102(172.30.200.102:3306) +--172.30.200.103(172.30.200.103:3306) Thu Jan 9 14:44:54 2020 - [info] Checking replication health on 172.30.200.102.. Thu Jan 9 14:44:54 2020 - [info] ok. Thu Jan 9 14:44:54 2020 - [info] Checking replication health on 172.30.200.103.. Thu Jan 9 14:44:54 2020 - [info] ok. Thu Jan 9 14:44:54 2020 - [warning] master_ip_failover_script is not defined. Thu Jan 9 14:44:54 2020 - [warning] shutdown_script is not defined. Thu Jan 9 14:44:54 2020 - [info] Got exit code 0 (Not master dead). MySQL Replication Health is OK.
3.2.6 MHA啟動
啟動mha manager:
nohup masterha_manager --conf=/etc/mha_master/mha.cnf &> /etc/mha_master/manager.log &檢測master節點狀態:
[root@localhost ~]# masterha_check_status --conf=/etc/mha_master/mha.cnf mha (pid:31709) is running(0:PING_OK), master:172.30.200.101說明主資料庫
172.30.200.101啟動正常。關閉mha manager:
masterha_stop -conf=/etc/mha_master/mha.cnf
3.2.7 MHA模擬故障
master直接kill mysql節點[root@localhost ~]# ps -ef |grep mysql root 19864 1 0 08:51 ? 00:00:00 /bin/sh /usr/local/mysql/bin/mysqld_safe --datadir=/data/mysqldata --pid-file=/data/mysqldata/localhost.localdomain.pid mysql 19976 19864 0 08:51 ? 00:00:13 /usr/local/mysql/bin/mysqld --basedir=/usr/local/mysql --datadir=/data/mysqldata --plugin-dir=/usr/local/mysql/lib/plugin --user=mysql --log-error=/data/mysqldata/mysqld.log --pid-file=/data/mysqldata/localhost.localdomain.pid --socket=/tmp/mysql.sock root 22166 21525 0 14:55 pts/0 00:00:00 grep --color=auto mysql [root@localhost ~]# kill -9 19864 19976MHA轉移日誌。[root@localhost ~]# tail -f /etc/mha_master/manager.log From: 172.30.200.101(172.30.200.101:3306) (current master) +--172.30.200.102(172.30.200.102:3306) +--172.30.200.103(172.30.200.103:3306) To: 172.30.200.102(172.30.200.102:3306) (new master) +--172.30.200.103(172.30.200.103:3306) Master 172.30.200.101(172.30.200.101:3306) is down! Check MHA Manager logs at localhost.localdomain:/etc/mha_master/manager.log for details. Started automated(non-interactive) failover. The latest slave 172.30.200.102(172.30.200.102:3306) has all relay logs for recovery. Selected 172.30.200.102(172.30.200.102:3306) as a new master. 172.30.200.102(172.30.200.102:3306): OK: Applying all logs succeeded. 172.30.200.103(172.30.200.103:3306): This host has the latest relay log events. Generating relay diff files from the latest slave succeeded. 172.30.200.103(172.30.200.103:3306): OK: Applying all logs succeeded. Slave started, replicating from 172.30.200.102(172.30.200.102:3306) 172.30.200.102(172.30.200.102:3306): Resetting slave info succeeded. Master failover to 172.30.200.102(172.30.200.102:3306) completed successfully.從上述日誌來看，
172.30.200.102已經成為了新的master，而172.30.200.103還是slave資料庫。
3.2.8 修復已損壞的節點加入MHA中
由於這裡是實驗環境，可以不到處mysqldump的備份。如果是生產環境恢復，可以停掉slave的SQL thread，記住對應的pos的位置，然後備份出資料，保證資料一致性之後，同步資料，恢復損壞的結點。
change master to master_host='172.30.200.102',
master_user='repl',
master_password='repl7101',
master_log_file='mysql-bin.000003',
master_log_pos=401;檢視slave狀態:
MariaDB [(none)]> start slave;
MariaDB [(none)]> show slave status\G;
*************************** 1. row ***************************
Slave_IO_State: Waiting for master to send event
Master_Host: 172.30.200.102
Master_User: repl
Master_Port: 3306
Connect_Retry: 60
Master_Log_File: mysql-bin.000003
Read_Master_Log_Pos: 401
Relay_Log_File: relay-bin.000002
Relay_Log_Pos: 555
Relay_Master_Log_File: mysql-bin.000003
Slave_IO_Running: Yes
Slave_SQL_Running: Yes再次啟動，如下:
[root@localhost ~]# nohup masterha_manager --conf=/etc/mha_master/mha.cnf &> /etc/mha_master/manager.log &
[root@localhost ~]# masterha_check_status --conf=/etc/mha_master/mha.cnf
mha (pid:31905) is running(0:PING_OK), master:172.30.200.101至此，MHA實驗完畢。由於生產環境會用到VIP，後續會繼續編寫。
MHA 問題集錦
錯誤一
日誌錯誤:
Thu Jan 9 11:31:36 2020 - [info] Connecting to [email protected](172.30.200.102:22)..
Can't exec "mysqlbinlog": 沒有那個檔案或目錄 at /usr/share/perl5/vendor_perl/MHA/BinlogManager.pm line 106.
mysqlbinlog version command failed with rc 1:0, please verify PATH, LD_LIBRARY_PATH, and client options解決方法:
ln -s /usr/local/mysql/bin/mysqlbinlog /usr/bin/mysqlbinlog錯誤二
日誌錯誤:
Checking if super_read_only is defined and turned on..DBI connect(';host=172.30.200.102;port=3306','mhaadmin',...) failed: Access denied for user 'mhaadmin'@'arpslave1' (using password: YES) at /usr/share/perl5/vendor_perl/MHA/SlaveUtil.pm line 239解決方法:
manager節點，執行:
grant all on *.* to 'mhaadmin'@'arpmgr' identified by 'mha7101';
grant all on *.* to 'mhaadmin'@'arpmaster' identified by 'mha7101';
grant all on *.* to 'mhaadmin'@'arpslave1' identified by 'mha7101';
grant all on *.* to 'mhaadmin'@'arpslave2' identified by 'mha7101';錯誤三
日誌如下:
Testing mysql connection and privileges..sh: mysql: 未找到命令
mysql command failed with rc 127:0!解決方法:
ln -s /usr/local/mysql/bin/mysql /usr/local/bin/mysql