体验MySQL MMM
?? ?今天試驗了一把MySQL MMM.（MySQL Master-Master Replication Manager）一個主主復制的管理工具，它提供了主機監控，MySQL服務監控，復制線程等多個方位的監控，使用VIP自動漂移，可以對應用程序實現透明。
?? ?下面是試驗步驟：
OS: Ubuntu 11.04 X64DB1:192.168.1.186 (MASTER1)DB2:192.168.1.187 (MASTER2)DB3:192.168.1.184 (SLAVE)Monitor: 192.168.1.175虛擬ip：192.168.1.200（writer），192.168.1.201/202（reader）?? ?MySQL V5.1.50（源碼安裝）
?? ?MMM V2.2.1 下載地址：http://mysql-mmm.org/downloads
?? ?安裝部分參考地址：參考地址：http://mysql-mmm.org/mmm2:guide
?? ?Note：在建立復制環境時需要注意的幾點：
1、server_id 要設置不同2、Auto_increment_offset=1 Auto_increment_increment=2 （master1 上設置）Auto_increment_offset=2 Auto_increment_increment=2 （master2上設置）?? ? ? 可以防止在某master假死并恢復的瞬間，兩個“master” 都出現數據的更新，導致復制失敗
?? ?3、master上都要啟用 log_slave_updates,log_bin
?? ?MMM 安裝：
?? ?所需用戶：
| monitor user | used by the mmm monitor to check the health of the MySQL servers | REPLICATION CLIENT |
| agent user | used by the mmm agent to change read-only mode, replication master, etc. | SUPER, REPLICATION CLIENT, PROCESS |
| relication user | used for replication | REPLICATION SLAVE ? |
?? ? perl 依賴包安裝：
?? ??（DB需要安裝部分：）
?? ?monitoring ?安裝部分：
perl -MCPAN -e shell cpan> install Algorithm::Diff cpan> install Class::Singleton cpan> install Log::Dispatch cpan> install Log::Log4perl cpan> install Mail::Send cpan> install Proc::Daemon cpan> install Thread::Queue cpan> install Time::HiRes cpan> install DBIcpan>install DBD::mysql?? ?下載的二進制包，直接make ?install (在所有機器上)
???? MMM主要的功能通過下面三個腳本來實現：
mmm_mond
監控進程，負責所有的監控工作，決定和處理所有節點角色活動
???? mmm_agentd
運行在每個mysql服務器上的代理進程，完成監控的探針工作和執行簡單的遠端服務設置
??? mmm_control
?一個簡單的腳本，提供管理mmm_mond進程的命令
?? ?配置文件部分：
active_master_role writer <host default>cluster_interface eth0pid_path /var/run/mmm_agentd.pidbin_path /usr/lib/mysql-mmm/replication_user abcreplication_password MMMagent_user mmm_agentagent_password MMMmysql_port 3308 </host> <host db1>ip 192.168.1.186mode masterpeer db2 </host> <host db2>ip 192.168.1.187mode masterpeer db1 </host> <host db3>ip 192.168.1.184mode slave </host> <role writer>hosts db2,db1ips 192.168.1.200mode exclusive </role> <role reader>hosts db2,db3ips 192.168.1.201,192.168.1.202mode balanced </role>?? ?monitoring 部分：
active_master_role writer <host default>cluster_interface eth0pid_path /var/run/mmm_agentd.pidbin_path /usr/lib/mysql-mmm/replication_user abcreplication_password MMMagent_user mmm_agentagent_password MMMmysql_port 3308 </host> <host db1>ip 192.168.1.186mode masterpeer db2 </host> <host db2>ip 192.168.1.187mode masterpeer db1 </host> <host db3>ip 192.168.1.184mode slave </host> <role writer>hosts db2,db1ips 192.168.1.200mode exclusive </role> <role reader>hosts db2,db3ips 192.168.1.201,192.168.1.202mode balanced </role> 安裝好后： mmm_control show 查看狀態： db1(192.168.1.186) master/ONLINE. Roles: writer(192.168.1.200)db2(192.168.1.187) master/ONLINE. Roles: reader(192.168.1.201)db3(192.168.1.184) slave/ONLINE. Roles: reader(192.168.1.202)?? ?測試一、db2 宕機，恢復情況：
?? ?日志分析：
2012/08/23 17:02:59 WARN Check 'rep_backlog' on 'db2' is in unknown state! Message: UNKNOWN: Connect error (host = 192.168.1.187:3308, user = mmm_monitor)! Lost connection to MySQL server at 'reading initial communication packet', system error: 1112012/08/23 17:02:59 WARN Check 'rep_threads' on 'db2' is in unknown state! Message: UNKNOWN: Connect error (host = 192.168.1.187:3308, user = mmm_monitor)! Lost connection to MySQL server at 'reading initial communication packet', system error: 1112012/08/23 17:03:09 ERROR Check 'mysql' on 'db2' has failed for 10 seconds! Message: ERROR: Connect error (host = 192.168.1.187:3308, user = mmm_monitor)! Lost connection to MySQL server at 'reading initial communication packet', system error: 1112012/08/23 17:03:09 ERROR Check 'rep_threads' on 'db1' has failed for 10 seconds! Message: ERROR: Replication is broken2012/08/23 17:03:10 FATAL State of host 'db2' changed from ONLINE to HARD_OFFLINE (ping: OK, mysql: not OK)2012/08/23 17:03:10 INFO Removing all roles from host 'db2':2012/08/23 17:03:10 INFO Removed role 'reader(192.168.1.201)' from host 'db2'2012/08/23 17:03:10 INFO Orphaned role 'reader(192.168.1.201)' has been assigned to 'db3'?? ?其中我們可以看到，monitor 通過10次 mysql檢查，判定DB2已經從 online狀態到?HARD_OFFLINE ?主機存在，但是數據庫已“死”，DB1的線程也已經停止，從 DB2上移除 reader角色到 DB3.
?? ?當把DB2啟動起來之后，它的狀態會變為：changed from HARD_OFFLINE to AWAITING_RECOVERY，
Check 'rep_threads' on 'db1' is ok!mmm_control set_online db2 手動變為online狀態，它會重新獲得reader角色：Admin changed state of 'db2' from AWAITING_RECOVERY to ONLINEINFO Moving role 'reader(192.168.1.202)' from host 'db3' to host 'db2'?? ?測試二：
?? ?DB1宕機， DB2應該變為writer，db3 slave 把主定向為db2.
?? ?日志分析：
2012/08/23 17:14:51 WARN Check 'rep_threads' on 'db1' is in unknown state! Message: UNKNOWN: Connect error (host = 192.168.1.186:3308, user = mmm_monitor)! Lost connection to MySQL server at 'reading initial communication packet', system error: 1112012/08/23 17:14:51 WARN Check 'rep_backlog' on 'db1' is in unknown state! Message: UNKNOWN: Connect error (host = 192.168.1.186:3308, user = mmm_monitor)! Lost connection to MySQL server at 'reading initial communication packet', system error: 1112012/08/23 17:15:00 ERROR Check 'mysql' on 'db1' has failed for 10 seconds! Message: ERROR: Connect error (host = 192.168.1.186:3308, user = mmm_monitor)! Lost connection to MySQL server at 'reading initial communication packet', system error: 1112012/08/23 17:15:01 ERROR Check 'rep_threads' on 'db2' has failed for 10 seconds! Message: ERROR: Replication is broken2012/08/23 17:15:01 ERROR Check 'rep_threads' on 'db3' has failed for 10 seconds! Message: ERROR: Replication is broken2012/08/23 17:15:02 FATAL State of host 'db1' changed from ONLINE to HARD_OFFLINE (ping: OK, mysql: not OK)2012/08/23 17:15:02 INFO Removing all roles from host 'db1':2012/08/23 17:15:02 INFO Removed role 'writer(192.168.1.200)' from host 'db1'2012/08/23 17:15:02 INFO Orphaned role 'writer(192.168.1.200)' has been assigned to 'db2'2012/08/23 17:15:06 FATAL State of host 'db3' changed from ONLINE to REPLICATION_FAIL2012/08/23 17:15:06 INFO Removing all roles from host 'db3':2012/08/23 17:15:06 INFO Removed role 'reader(192.168.1.201)' from host 'db3'2012/08/23 17:15:06 INFO Check 'rep_threads' on 'db3' is ok!2012/08/23 17:15:06 INFO Orphaned role 'reader(192.168.1.201)' has been assigned to 'db2'2012/08/23 17:15:09 FATAL State of host 'db3' changed from REPLICATION_FAIL to ONLINE2012/08/23 17:15:09 INFO Moving role 'reader(192.168.1.202)' from host 'db2' to host 'db3'?? ?Monitor 檢測到 DB2和DB3 與DB1 之間的復制已經停止，10次檢查后，確定DB1 已經 “死掉”，將它的writer ip 轉移至 DB2，db3 復制線程停止后，將其擁有的 reader ip摘除，當 db3 重新與 DB2建立主從關系后，又獲得 reader 角色；
?? ?等DB1恢復后， ?mmm_control show
db1(192.168.1.186) master/ONLINE. Roles: db2(192.168.1.187) master/ONLINE. Roles: reader(192.168.1.201), writer(192.168.1.200)db3(192.168.1.184) slave/ONLINE. Roles: reader(192.168.1.202)?? 手動將DB1獲得writer角色：
mmm_control move_role writer db1db1(192.168.1.186) master/ONLINE. Roles: writer(192.168.1.200)db2(192.168.1.187) master/ONLINE. Roles: reader(192.168.1.201)db3(192.168.1.184) slave/ONLINE. Roles: reader(192.168.1.202)?? 在測試的時候，間隔要超過60s，原因是：MMM防抖動檢測：
?? 當主機從Online 狀態轉換為HARD_OFFLINE/REPLICATION_FAIL/REPLICATION_DELAY，又從這些狀態轉換為online（在小于60s的情況下）
?? ?MMM 會阻止這種情況的發生，我們可以通過配置文件配置它，如果一個主機在flap_duration時間內宕掉了flap_count次，就認為主機處理flap狀態，這個主機就不會自動被設置為ONLINE狀 態，它將一直處于AWAITING_RECOVERY狀態除非手動設置online(mmm_control set online host)。如果auto_set_online>0,處于flapping的主機在flap_duration時間后將制動被設置為ONLINE 狀態。
?? 英文參考地址：http://mysql-mmm.org/mysql-mmm.html
?? 中文參考地址：http://linuxguest.blog.51cto.com/195664/578311
?? ?
?
?
?
?
?? ??
轉載于:https://blog.51cto.com/weipengfei/971419
總結
以上是生活随笔為你收集整理的体验MySQL MMM的全部內容，希望文章能夠幫你解決所遇到的問題。
- 上一篇： System Center 2012与v
- 下一篇： linux cmake编译源码,linu