mysqlPXC集群脑裂及grastate.dat修改实验
三台服务器做了 MySQL PXC 集群
创新互联公司是一家专注网站建设、网络营销策划、微信小程序开发、电子商务建设、网络推广、移动互联开发、研究、服务为一体的技术型公司。公司成立十年以来，已经为近1000家成都酒楼设计各业的企业公司提供互联网服务。现在，服务的近1000家客户与我们一路同行，见证我们的成长；未来，我们一起分享成功的喜悦。
172.31.217.182 bd-dev-mingshuo-182
172.31.217.183 bd-dev-mingshuo-183
172.31.217.89 bd-dev-vertica-89
正常关闭 183 一个节点
mysqladmin -uroot -poracle -S /u01/mysql/3307/data/mysql.sock -P3307 shutdown
关闭节点 log ：
2018-09-27T07:33:13.222079Z 0 [Note] WSREP: Received shutdown signal. Will sleep for 10 secs before initiating shutdown. pxc_maint_mode switched to SHUTDOWN 2018-09-27T07:33:23.230509Z 0 [Note] WSREP: Stop replication 2018-09-27T07:33:23.230619Z 0 [Note] WSREP: Closing send monitor... 2018-09-27T07:33:23.230640Z 0 [Note] WSREP: Closed send monitor. 2018-09-27T07:33:23.230660Z 0 [Note] WSREP: gcomm: terminating thread 2018-09-27T07:33:23.230680Z 0 [Note] WSREP: gcomm: joining thread 2018-09-27T07:33:23.230827Z 0 [Note] WSREP: gcomm: closing backend 2018-09-27T07:33:23.231780Z 0 [Note] WSREP: Current view of cluster as seen by this node view (view_id(NON_PRIM,12f1e199,11) memb { 12f1e199,0 } joined { } left { } partitioned { 2331d3d7,0 c05737fd,0 } ) 2018-09-27T07:33:23.231867Z 0 [Note] WSREP: Current view of cluster as seen by this node view ((empty)) 2018-09-27T07:33:23.232111Z 0 [Note] WSREP: gcomm: closed 2018-09-27T07:33:23.232165Z 0 [Note] WSREP: New COMPONENT: primary = no, bootstrap = no, my_idx = 0, memb_num = 1 2018-09-27T07:33:23.232253Z 0 [Note] WSREP: Flow-control interval: [100, 100] 2018-09-27T07:33:23.232260Z 0 [Note] WSREP: Trying to continue unpaused monitor 2018-09-27T07:33:23.232264Z 0 [Note] WSREP: Received NON-PRIMARY. 2018-09-27T07:33:23.232268Z 0 [Note] WSREP: Shifting SYNCED -> OPEN (TO: 27) 2018-09-27T07:33:23.232279Z 0 [Note] WSREP: Received self-leave message. 2018-09-27T07:33:23.232285Z 0 [Note] WSREP: Flow-control interval: [0, 0] 2018-09-27T07:33:23.232288Z 0 [Note] WSREP: Trying to continue unpaused monitor 2018-09-27T07:33:23.232291Z 0 [Note] WSREP: Received SELF-LEAVE. Closing connection. 2018-09-27T07:33:23.232295Z 0 [Note] WSREP: Shifting OPEN -> CLOSED (TO: 27) 2018-09-27T07:33:23.232302Z 0 [Note] WSREP: RECV thread exiting 0: Success 2018-09-27T07:33:23.232383Z 2 [Note] WSREP: New cluster view: global state: c057dbc5-c16e-11e8-a1a6-825ed9079934:27, view# -1: non-Primary, number of nodes: 1, my index: 0, protocol version 3 2018-09-27T07:33:23.232394Z 2 [Note] WSREP: Setting wsrep_ready to false 2018-09-27T07:33:23.232400Z 2 [Note] WSREP: wsrep_notify_cmd is not defined, skipping notification. 2018-09-27T07:33:23.232439Z 2 [Note] WSREP: New cluster view: global state: c057dbc5-c16e-11e8-a1a6-825ed9079934:27, view# -1: non-Primary, number of nodes: 0, my index: -1, protocol version 3 2018-09-27T07:33:23.232443Z 2 [Note] WSREP: Setting wsrep_ready to false 2018-09-27T07:33:23.232446Z 2 [Note] WSREP: wsrep_notify_cmd is not defined, skipping notification. 2018-09-27T07:33:23.232472Z 2 [Note] WSREP: applier thread exiting (code:0) 2018-09-27T07:33:23.232479Z 0 [Note] WSREP: recv_thread() joined. 2018-09-27T07:33:23.232502Z 0 [Note] WSREP: Closing replication queue. 2018-09-27T07:33:23.232509Z 0 [Note] WSREP: Closing slave action queue. 2018-09-27T07:33:23.232517Z 0 [Note] Giving 2 client threads a chance to die gracefully 2018-09-27T07:33:25.232639Z 0 [Note] WSREP: Waiting for active wsrep applier to exit 2018-09-27T07:33:25.232758Z 1 [Note] WSREP: rollbacker thread exiting 2018-09-27T07:33:25.232994Z 0 [Note] Giving 0 client threads a chance to die gracefully 2018-09-27T07:33:25.233010Z 0 [Note] Shutting down slave threads 2018-09-27T07:33:25.233025Z 0 [Note] Forcefully disconnecting 0 remaining clients 2018-09-27T07:33:25.233044Z 0 [Note] Event Scheduler: Purging the queue. 0 events 2018-09-27T07:33:25.242788Z 0 [Note] WSREP: Service thread queue flushed. 2018-09-27T07:33:25.250399Z 0 [Note] WSREP: MemPool(SlaveTrxHandle): hit ratio: 0, misses: 0, in use: 0, in pool: 0 2018-09-27T07:33:25.250479Z 0 [Note] WSREP: Shifting CLOSED -> DESTROYED (TO: 27) 2018-09-27T07:33:25.259428Z 0 [Note] Binlog end 2018-09-27T07:33:25.261702Z 0 [Note] Shutting down plugin 'ngram' 2018-09-27T07:33:25.261721Z 0 [Note] Shutting down plugin 'partition' 2018-09-27T07:33:25.261726Z 0 [Note] Shutting down plugin 'ARCHIVE' 2018-09-27T07:33:25.261729Z 0 [Note] Shutting down plugin 'BLACKHOLE' 2018-09-27T07:33:25.261733Z 0 [Note] Shutting down plugin 'INNODB_SYS_VIRTUAL' 2018-09-27T07:33:25.261736Z 0 [Note] Shutting down plugin 'INNODB_CHANGED_PAGES' 2018-09-27T07:33:25.261739Z 0 [Note] Shutting down plugin 'INNODB_SYS_DATAFILES' 2018-09-27T07:33:25.261741Z 0 [Note] Shutting down plugin 'INNODB_SYS_TABLESPACES' 2018-09-27T07:33:25.261744Z 0 [Note] Shutting down plugin 'INNODB_SYS_FOREIGN_COLS' 2018-09-27T07:33:25.261746Z 0 [Note] Shutting down plugin 'INNODB_SYS_FOREIGN' 2018-09-27T07:33:25.261749Z 0 [Note] Shutting down plugin 'INNODB_SYS_FIELDS' 2018-09-27T07:33:25.261751Z 0 [Note] Shutting down plugin 'INNODB_SYS_COLUMNS' 2018-09-27T07:33:25.261754Z 0 [Note] Shutting down plugin 'INNODB_SYS_INDEXES' 2018-09-27T07:33:25.261756Z 0 [Note] Shutting down plugin 'INNODB_SYS_TABLESTATS' 2018-09-27T07:33:25.261759Z 0 [Note] Shutting down plugin 'INNODB_SYS_TABLES' 2018-09-27T07:33:25.261761Z 0 [Note] Shutting down plugin 'INNODB_FT_INDEX_TABLE' 2018-09-27T07:33:25.261764Z 0 [Note] Shutting down plugin 'INNODB_FT_INDEX_CACHE' 2018-09-27T07:33:25.261766Z 0 [Note] Shutting down plugin 'INNODB_FT_CONFIG' 2018-09-27T07:33:25.261769Z 0 [Note] Shutting down plugin 'INNODB_FT_BEING_DELETED' 2018-09-27T07:33:25.261771Z 0 [Note] Shutting down plugin 'INNODB_FT_DELETED' 2018-09-27T07:33:25.261774Z 0 [Note] Shutting down plugin 'INNODB_FT_DEFAULT_STOPWORD' 2018-09-27T07:33:25.261776Z 0 [Note] Shutting down plugin 'INNODB_METRICS' 2018-09-27T07:33:25.261778Z 0 [Note] Shutting down plugin 'INNODB_TEMP_TABLE_INFO' 2018-09-27T07:33:25.261781Z 0 [Note] Shutting down plugin 'INNODB_BUFFER_POOL_STATS' 2018-09-27T07:33:25.261783Z 0 [Note] Shutting down plugin 'INNODB_BUFFER_PAGE_LRU' 2018-09-27T07:33:25.261785Z 0 [Note] Shutting down plugin 'INNODB_BUFFER_PAGE' 2018-09-27T07:33:25.261788Z 0 [Note] Shutting down plugin 'INNODB_CMP_PER_INDEX_RESET' 2018-09-27T07:33:25.261790Z 0 [Note] Shutting down plugin 'INNODB_CMP_PER_INDEX' 2018-09-27T07:33:25.261793Z 0 [Note] Shutting down plugin 'INNODB_CMPMEM_RESET' 2018-09-27T07:33:25.261795Z 0 [Note] Shutting down plugin 'INNODB_CMPMEM' 2018-09-27T07:33:25.261797Z 0 [Note] Shutting down plugin 'INNODB_CMP_RESET' 2018-09-27T07:33:25.261800Z 0 [Note] Shutting down plugin 'INNODB_CMP' 2018-09-27T07:33:25.261802Z 0 [Note] Shutting down plugin 'INNODB_LOCK_WAITS' 2018-09-27T07:33:25.261805Z 0 [Note] Shutting down plugin 'INNODB_LOCKS' 2018-09-27T07:33:25.261807Z 0 [Note] Shutting down plugin 'INNODB_TRX' 2018-09-27T07:33:25.261809Z 0 [Note] Shutting down plugin 'XTRADB_ZIP_DICT_COLS' 2018-09-27T07:33:25.261812Z 0 [Note] Shutting down plugin 'XTRADB_ZIP_DICT' 2018-09-27T07:33:25.261814Z 0 [Note] Shutting down plugin 'XTRADB_RSEG' 2018-09-27T07:33:25.261817Z 0 [Note] Shutting down plugin 'XTRADB_INTERNAL_HASH_TABLES' 2018-09-27T07:33:25.261819Z 0 [Note] Shutting down plugin 'XTRADB_READ_VIEW' 2018-09-27T07:33:25.261822Z 0 [Note] Shutting down plugin 'InnoDB' 2018-09-27T07:33:25.261857Z 0 [Note] InnoDB: FTS optimize thread exiting. 2018-09-27T07:33:25.262097Z 0 [Note] InnoDB: Starting shutdown... 2018-09-27T07:33:25.362428Z 0 [Note] InnoDB: Dumping buffer pool(s) to /u01/mysql/3307/data/ib_buffer_pool 2018-09-27T07:33:25.363022Z 0 [Note] InnoDB: Buffer pool(s) dump completed at 180927 15:33:25 2018-09-27T07:33:25.562786Z 0 [Note] InnoDB: Waiting for page_cleaner to finish flushing of buffer pool 2018-09-27T07:33:26.571050Z 0 [Note] InnoDB: Shutdown completed; log sequence number 2569669 2018-09-27T07:33:26.574169Z 0 [Note] InnoDB: Removed temporary tablespace data file: "ibtmp1" 2018-09-27T07:33:26.574193Z 0 [Note] Shutting down plugin 'MyISAM' 2018-09-27T07:33:26.574210Z 0 [Note] Shutting down plugin 'MRG_MYISAM' 2018-09-27T07:33:26.574222Z 0 [Note] Shutting down plugin 'CSV' 2018-09-27T07:33:26.574233Z 0 [Note] Shutting down plugin 'MEMORY' 2018-09-27T07:33:26.574254Z 0 [Note] Shutting down plugin 'PERFORMANCE_SCHEMA' 2018-09-27T07:33:26.574287Z 0 [Note] Shutting down plugin 'sha256_password' 2018-09-27T07:33:26.574296Z 0 [Note] Shutting down plugin 'mysql_native_password' 2018-09-27T07:33:26.574304Z 0 [Note] Shutting down plugin 'wsrep' 2018-09-27T07:33:26.574480Z 0 [Note] Shutting down plugin 'binlog' |
正常节点 log
2018-09-27T07:33:22.505216Z 0 [Note] WSREP: declaring c05737fd at tcp://172.31.217.89:4567 stable 2018-09-27T07:33:22.505345Z 0 [Note] WSREP: forgetting 12f1e199 (tcp://172.31.217.183:4567) 2018-09-27T07:33:22.511586Z 0 [Note] WSREP: Node 2331d3d7 state primary 2018-09-27T07:33:22.512245Z 0 [Note] WSREP: Current view of cluster as seen by this node view (view_id(PRIM,2331d3d7,12) memb { 2331d3d7,0 c05737fd,0 } joined { } left { } partitioned { 12f1e199,0 } ) 2018-09-27T07:33:22.512303Z 0 [Note] WSREP: Save the discovered primary-component to disk 2018-09-27T07:33:22.512547Z 0 [Note] WSREP: New COMPONENT: primary = yes, bootstrap = no, my_idx = 0, memb_num = 2 2018-09-27T07:33:22.513157Z 0 [Note] WSREP: forgetting 12f1e199 (tcp://172.31.217.183:4567) 2018-09-27T07:33:22.513241Z 0 [Note] WSREP: STATE_EXCHANGE: sent state UUID: 9ccf0351-c227-11e8-ae6a-d3cac5b411a7 2018-09-27T07:33:22.514096Z 0 [Note] WSREP: STATE EXCHANGE: sent state msg: 9ccf0351-c227-11e8-ae6a-d3cac5b411a7 2018-09-27T07:33:22.514647Z 0 [Note] WSREP: STATE EXCHANGE: got state msg: 9ccf0351-c227-11e8-ae6a-d3cac5b411a7 from 0 (bd-dev-mingshuo-182) 2018-09-27T07:33:22.514661Z 0 [Note] WSREP: STATE EXCHANGE: got state msg: 9ccf0351-c227-11e8-ae6a-d3cac5b411a7 from 1 (bd-dev-vertica-89) 2018-09-27T07:33:22.514669Z 0 [Note] WSREP: Quorum results: version = 4, component = PRIMARY, conf_id = 11, members = 2/2 (primary/total), act_id = 27, last_appl. = 0, protocols = 0/8/3 (gcs/repl/appl), group UUID = c057dbc5-c16e-11e8-a1a6-825ed9079934 2018-09-27T07:33:22.514675Z 0 [Note] WSREP: Flow-control interval: [141, 141] 2018-09-27T07:33:22.514679Z 0 [Note] WSREP: Trying to continue unpaused monitor 2018-09-27T07:33:22.514707Z 2 [Note] WSREP: New cluster view: global state: c057dbc5-c16e-11e8-a1a6-825ed9079934:27, view# 12: Primary, number of nodes: 2, my index: 0, protocol version 3 2018-09-27T07:33:22.514713Z 2 [Note] WSREP: Setting wsrep_ready to true 2018-09-27T07:33:22.514719Z 2 [Note] WSREP: wsrep_notify_cmd is not defined, skipping notification. 2018-09-27T07:33:22.514727Z 2 [Note] WSREP: REPL Protocols: 8 (3, 2) 2018-09-27T07:33:22.514747Z 2 [Note] WSREP: Assign initial position for certification: 27, protocol version: 3 2018-09-27T07:33:22.514830Z 0 [Note] WSREP: Service thread queue flushed. 2018-09-27T07:33:27.691129Z 0 [Note] WSREP: cleaning up 12f1e199 (tcp://172.31.217.183:4567) |
182 节点插入数据
mysql> insert into t1 values (4,4); Query OK, 1 row affected (0.01 sec)
mysql> select * from t1; +---+------+ | a | b | +---+------+ | 1 | 1 | | 2 | 2 | | 3 | 3 | | 4 | 4 | +---+------+ 4 rows in set (0.00 sec) |
启动 183 节点
mysql -S /u01/mysql/3307/data/mysql.sock -uroot -poracle -P3307 mysql> select * from t1; +---+------+ | a | b | +---+------+ | 1 | 1 | | 2 | 2 | | 3 | 3 | | 4 | 4 | +---+------+ 4 rows in set (0.00 sec) |
增量数据已经同步过来了。
下面是日志增量应用部分，可以看到收到了一个事务。
2018-09-27T08:05:50.785769Z 0 [Note] WSREP: Signalling provider to continue on SST completion. 2018-09-27T08:05:50.785808Z 0 [Note] WSREP: Initialized wsrep sidno 2 2018-09-27T08:05:50.785833Z 0 [Note] WSREP: SST received: c057dbc5-c16e-11e8-a1a6-825ed9079934:27 2018-09-27T08:05:50.785872Z 2 [Note] WSREP: Receiving IST: 1 writesets, seqnos 27-28 2018-09-27T08:05:50.785985Z 0 [Note] 2018-09-27T08:05:50.785985Z 0 [Note] WSREP: Receiving IST... 0.0% (0/1 events) complete. 2018-09-27T08:05:50.877679Z 0 [Note] WSREP: Receiving IST...100.0% (1/1 events) complete. 2018-09-27T08:05:50.877904Z 2 [Note] WSREP: IST received: c057dbc5-c16e-11e8-a1a6-825ed9079934:28 2018-09-27T08:05:50.878589Z 0 [Note] WSREP: 1.0 (bd-dev-mingshuo-183): State transfer from 0.0 (bd-dev-mingshuo-182) complete. 2018-09-27T08:05:50.878603Z 0 [Note] WSREP: SST leaving flow control 2018-09-27T08:05:50.878608Z 0 [Note] WSREP: Shifting JOINER -> JOINED (TO: 28) 2018-09-27T08:05:50.879059Z 0 [Note] WSREP: Member 1.0 (bd-dev-mingshuo-183) synced with group. 2018-09-27T08:05:50.879072Z 0 [Note] WSREP: Shifting JOINED -> SYNCED (TO: 28) 2018-09-27T08:05:50.879101Z 2 [Note] WSREP: Synchronized with group, ready for connections 2018-09-27T08:05:50.879115Z 2 [Note] WSREP: Setting wsrep_ready to true |
现在测试直接非正常关闭两个节点
将 182 和 183 两个节点进程直接 kill -9 杀掉
89 存活节点：
mysql> select * from ming.t1; ERROR 1047 (08S01): WSREP has not yet prepared node for application use mysql> insert into ming.t1 values(10,10); ERROR 1047 (08S01): WSREP has not yet prepared node for application use |
存活节点已经无法正常提供读写服务。
mysql> show status where Variable_name IN ('wsrep_local_state_uuid','wsrep_cluster_conf_id','wsrep_cluster_size', 'wsrep_cluster_status','wsrep_ready','wsrep_connected'); +------------------------+--------------------------------------+ | Variable_name | Value | +------------------------+--------------------------------------+ | wsrep_local_state_uuid | c057dbc5-c16e-11e8-a1a6-825ed9079934 | | wsrep_cluster_conf_id | 18446744073709551615 | | wsrep_cluster_size | 1 | | wsrep_cluster_status | non-Primary | | wsrep_connected | ON | | wsrep_ready | OFF | +------------------------+--------------------------------------+ 6 rows in set (0.00 sec) |
可以看到 wsrep_cluster_size=1 代表集群节点个数只剩自己了。
wsrep_cluster_status=non-Primary 代表集群状态不一致
wsrep_connected=ON 代表数据库还接受连接
wsrep_read=OFF 代表数据库已经不能正常接受查询服务了。上面的 select 语句也佐证了这一点。
存活节点能否提供读服务，取决于 wsrep_dirty_reads 参数
mysql> show variables like 'wsrep_dirty_reads'; +-------------------+-------+ | Variable_name | Value | +-------------------+-------+ | wsrep_dirty_reads | OFF | +-------------------+-------+ 1 row in set (0.00 sec) |
wsrep_dirty_reads 是可以动态调整的。如果设置为 ON ，那么在节点状态是 non-Primary 时，
是可以提供读的服务的。写的服务还需要提升该节点为 primary ，这个是要通过其他参数设定的，
后面会说的。
存活节点一直在尝试连接另外两个节点
2018-09-28T02:57:37.209095Z 0 [Note] WSREP: (725136c0, 'tcp://0.0.0.0:4567') reconnecting to 2331d3d7 (tcp://172.31.217.182:4567), attempt 960 2018-09-28T02:58:12.714612Z 0 [Note] WSREP: (725136c0, 'tcp://0.0.0.0:4567') reconnecting to 252da778 (tcp://172.31.217.183:4567), attempt 900 2018-09-28T02:58:22.216078Z 0 [Note] WSREP: (725136c0, 'tcp://0.0.0.0:4567') reconnecting to 2331d3d7 (tcp://172.31.217.182:4567), attempt 990 2018-09-28T02:58:57.721850Z 0 [Note] WSREP: (725136c0, 'tcp://0.0.0.0:4567') reconnecting to 252da778 (tcp://172.31.217.183:4567), attempt 930 2018-09-28T02:59:07.223430Z 0 [Note] WSREP: (725136c0, 'tcp://0.0.0.0:4567') reconnecting to 2331d3d7 (tcp://172.31.217.182:4567), attempt 1020 |
不能提供读写的原因其实就是 PXC 对集群脑裂的判断机制还不完善，对我自己来说我是 kill 掉了两个节点的进程。
但是对 PXC 来说，存活节点不知道另外两个节点的状态，有可能另外两个节点已经死掉了，有可能另外两个节点相互之间还能继续通信对外提供服务，
这样一来就形成了两个信息孤岛，彼此之间不能联系对方，所以存活节点就变成了这样不能读写的状态。
拉起两个节点后存活节点日志
2018-09-28T03:04:57.003914Z 0 [Note] WSREP: (725136c0, 'tcp://0.0.0.0:4567') connection established to 252da778 tcp://172.31.217.183:4567 2018-09-28T03:05:03.215507Z 0 [Note] WSREP: declaring 252da778 at tcp://172.31.217.183:4567 stable 2018-09-28T03:05:03.216346Z 0 [Note] WSREP: Current view of cluster as seen by this node view (view_id(NON_PRIM,252da778,30) memb { 252da778,0 725136c0,0 } joined { } left { } partitioned { 2331d3d7,0 } ) 2018-09-28T03:05:03.216630Z 0 [Note] WSREP: New COMPONENT: primary = no, bootstrap = no, my_idx = 1, memb_num = 2 2018-09-28T03:05:03.216710Z 0 [Note] WSREP: Flow-control interval: [141, 141] 2018-09-28T03:05:03.216718Z 0 [Note] WSREP: Trying to continue unpaused monitor 2018-09-28T03:05:03.216723Z 0 [Note] WSREP: Received NON-PRIMARY. 2018-09-28T03:05:03.216794Z 1 [Note] WSREP: New cluster view: global state: c057dbc5-c16e-11e8-a1a6-825ed9079934:33, view# -1: non-Primary, number of nodes: 2, my index: 1, protocol version 3 2018-09-28T03:05:03.216822Z 1 [Note] WSREP: Setting wsrep_ready to false 2018-09-28T03:05:03.216833Z 1 [Note] WSREP: wsrep_notify_cmd is not defined, skipping notification. 2018-09-28T03:05:04.277523Z 0 [Note] WSREP: (725136c0, 'tcp://0.0.0.0:4567') connection established to 2331d3d7 tcp://172.31.217.182:4567 2018-09-28T03:05:04.279018Z 0 [Note] WSREP: (725136c0, 'tcp://0.0.0.0:4567') connection established to 2331d3d7 tcp://172.31.217.182:4567 2018-09-28T03:05:04.776965Z 0 [Note] WSREP: declaring 2331d3d7 at tcp://172.31.217.182:4567 stable 2018-09-28T03:05:04.777019Z 0 [Note] WSREP: declaring 252da778 at tcp://172.31.217.183:4567 stable 2018-09-28T03:05:04.777487Z 0 [Note] WSREP: re-bootstrapping prim from partitioned components 2018-09-28T03:05:04.778262Z 0 [Note] WSREP: Current view of cluster as seen by this node view (view_id(PRIM,2331d3d7,31) memb { 2331d3d7,0 252da778,0 725136c0,0 } joined { } left { } partitioned { } ) 2018-09-28T03:05:04.778307Z 0 [Note] WSREP: Save the discovered primary-component to disk 2018-09-28T03:05:04.778588Z 0 [Note] WSREP: New COMPONENT: primary = yes, bootstrap = no, my_idx = 2, memb_num = 3 2018-09-28T03:05:04.778629Z 0 [Note] WSREP: STATE EXCHANGE: Waiting for state UUID. 2018-09-28T03:05:05.277931Z 0 [Note] WSREP: STATE EXCHANGE: sent state msg: 838d4806-c2cb-11e8-8bb1-eeeae1741165 2018-09-28T03:05:05.278435Z 0 [Note] WSREP: STATE EXCHANGE: got state msg: 838d4806-c2cb-11e8-8bb1-eeeae1741165 from 0 (bd-dev-mingshuo-182) 2018-09-28T03:05:05.278463Z 0 [Note] WSREP: STATE EXCHANGE: got state msg: 838d4806-c2cb-11e8-8bb1-eeeae1741165 from 1 (bd-dev-mingshuo-183) 2018-09-28T03:05:05.278470Z 0 [Note] WSREP: STATE EXCHANGE: got state msg: 838d4806-c2cb-11e8-8bb1-eeeae1741165 from 2 (bd-dev-vertica-89) 2018-09-28T03:05:05.278490Z 0 [Warning] WSREP: Quorum: No node with complete state:
Version : 4 Flags : 0x1 Protocols : 0 / 8 / 3 State : NON-PRIMARY Desync count : 0 Prim state : NON-PRIMARY Prim UUID : 00000000-0000-0000-0000-000000000000 Prim seqno : -1 First seqno : -1 Last seqno : 33 Prim JOINED : 0 State UUID : 838d4806-c2cb-11e8-8bb1-eeeae1741165 Group UUID : c057dbc5-c16e-11e8-a1a6-825ed9079934 Name : 'bd-dev-mingshuo-182' Incoming addr: '172.31.217.182:3307'
Version : 4 Flags : 00 Protocols : 0 / 8 / 3 State : NON-PRIMARY Desync count : 0 Prim state : NON-PRIMARY Prim UUID : 00000000-0000-0000-0000-000000000000 Prim seqno : -1 First seqno : -1 Last seqno : 33 Prim JOINED : 0 State UUID : 838d4806-c2cb-11e8-8bb1-eeeae1741165 Group UUID : c057dbc5-c16e-11e8-a1a6-825ed9079934 Name : 'bd-dev-mingshuo-183' Incoming addr: '172.31.217.183:3307'
Version : 4 Flags : 0x2 Protocols : 0 / 8 / 3 State : NON-PRIMARY Desync count : 0 Prim state : SYNCED Prim UUID : 19faf204-c2c7-11e8-b642-52dd65ccae43 Prim seqno : 26 First seqno : 33 Last seqno : 33 Prim JOINED : 2 State UUID : 838d4806-c2cb-11e8-8bb1-eeeae1741165 Group UUID : c057dbc5-c16e-11e8-a1a6-825ed9079934 Name : 'bd-dev-vertica-89' Incoming addr: '172.31.217.89:3307'
2018-09-28T03:05:05.278511Z 0 [Note] WSREP: Partial re-merge of primary 19faf204-c2c7-11e8-b642-52dd65ccae43 found: 1 of 2. 2018-09-28T03:05:05.278520Z 0 [Note] WSREP: Quorum results: version = 4, component = PRIMARY, conf_id = 26, members = 3/3 (primary/total), act_id = 33, last_appl. = 0, protocols = 0/8/3 (gcs/repl/appl), group UUID = c057dbc5-c16e-11e8-a1a6-825ed9079934 2018-09-28T03:05:05.278540Z 0 [Note] WSREP: Flow-control interval: [173, 173] 2018-09-28T03:05:05.278544Z 0 [Note] WSREP: Trying to continue unpaused monitor 2018-09-28T03:05:05.278548Z 0 [Note] WSREP: Restored state OPEN -> SYNCED (33) 2018-09-28T03:05:05.278593Z 1 [Note] WSREP: New cluster view: global state: c057dbc5-c16e-11e8-a1a6-825ed9079934:33, view# 27: Primary, number of nodes: 3, my index: 2, protocol version 3 2018-09-28T03:05:05.278612Z 1 [Note] WSREP: Setting wsrep_ready to true 2018-09-28T03:05:05.278621Z 1 [Note] WSREP: wsrep_notify_cmd is not defined, skipping notification. 2018-09-28T03:05:05.278661Z 1 [Note] WSREP: REPL Protocols: 8 (3, 2) 2018-09-28T03:05:05.278679Z 1 [Note] WSREP: Assign initial position for certification: 33, protocol version: 3 2018-09-28T03:05:05.278752Z 0 [Note] WSREP: Service thread queue flushed. 2018-09-28T03:05:05.278828Z 1 [Note] WSREP: Synchronized with group, ready for connections 2018-09-28T03:05:05.278849Z 1 [Note] WSREP: Setting wsrep_ready to true 2018-09-28T03:05:05.278863Z 1 [Note] WSREP: wsrep_notify_cmd is not defined, skipping notification. 2018-09-28T03:05:05.279134Z 0 [Note] WSREP: Member 0.0 (bd-dev-mingshuo-182) synced with group. 2018-09-28T03:05:05.279179Z 0 [Note] WSREP: Member 1.0 (bd-dev-mingshuo-183) synced with group. 2018-09-28T03:05:07.290328Z 0 [Note] WSREP: (725136c0, 'tcp://0.0.0.0:4567') turning message relay requesting off |
出现脑裂后解决方法：
SET GLOBAL wsrep_provider_options='pc.bootstrap=YES'; |
三个节点正常关闭
依次关闭 183,182,89 三个节点
启动的之前，一定要看一下 grastate.dat 文件内容
183 节点
root@bd-dev-mingshuo-183:/u01/mysql/3307/data#more grastate.dat # GALERA saved state version: 2.1 uuid: c057dbc5-c16e-11e8-a1a6-825ed9079934 seqno: 51 safe_to_bootstrap: 0
|
182 节点
root@bd-dev-mingshuo-182:/opt/mysql/3307/data#more grastate.dat # GALERA saved state version: 2.1 uuid: c057dbc5-c16e-11e8-a1a6-825ed9079934 seqno: 51 safe_to_bootstrap: 0 |
89 节点
root@bd-dev-vertica-89:/opt/mysql/3307/data#more grastate.dat # GALERA saved state version: 2.1 uuid: c057dbc5-c16e-11e8-a1a6-825ed9079934 seqno: 51 safe_to_bootstrap: 1 |
注意：
safe_to_bootstrap=1 的节点，说明这个节点是可以安全的作为主节点启动的。所以启动的时候必须先启动 89 节点。
mysqld_safe --defaults-file=/etc/my.cnf --wsrep-new-cluster &
mysqld_safe --defaults-file=/etc/my3307.cnf &
mysqld_safe --defaults-file=/etc/my3307.cnf &
疑问： mysql PXC 在启动时是不是只是按照 grastate.dat 的 safe_to_bootstrap 来验证集群呢？
这个很好证明，还是按照上面的做法关闭集群，然后修改 183 的 grastate.dat 的 safe_to_bootstrap 值为 1.
实验过程省略，但是这样做确实是可以启动集群的。
如果在关闭部分节点后有数据变化呢？
关闭 183,182 节点后，在 89 节点插入数据
mysql> insert into ming.t1 values (16,16); Query OK, 1 row affected (0.01 sec)
|
然后关闭 89 节点。至此集群全部关闭。
修改 183 的节点的 grastate.dat
root@bd-dev-mingshuo-183:/u01/mysql/3307/data#more grastate.dat # GALERA saved state version: 2.1 uuid: c057dbc5-c16e-11e8-a1a6-825ed9079934 seqno: 51 safe_to_bootstrap: 1 |
启动集群，先启动 183 节点：
mysqld_safe --defaults-file=/etc/my3307.cnf --wsrep-new-cluster & mysql> select * from ming.t1; +----+------+ | a | b | +----+------+ | 1 | 1 | | 2 | 2 | | 3 | 3 | | 4 | 4 | | 5 | 5 | | 6 | 6 | | 7 | 7 | | 8 | 8 | | 9 | 9 | | 10 | 10 | | 11 | 11 | | 12 | 12 | | 13 | 13 | | 14 | 14 | | 15 | 15 | +----+------+ 15 rows in set (0.00 sec) |
16 那行数据丢失了。
再去启动 89 节点，看看丢失的数据能否找回来
2018-09-28T08:13:33.156722Z 0 [ERROR] WSREP: gcs/src/gcs_group.cpp:group_post_state_exchange():322: Reversing history: 52 -> 51, this member has applied 1 more events than the primary component.Data loss is possible. Aborting. |
89 节点的日志序列已经到了 52 ，超过了其他节点的 51. 修改 89 节点日志序列为 51 ，然后再尝试启动 89 节点
mysqld_safe --defaults-file=/etc/my.cnf &
mysql> select * from ming.t1; +----+------+ | a | b | +----+------+ | 1 | 1 | | 2 | 2 | | 3 | 3 | | 4 | 4 | | 5 | 5 | | 6 | 6 | | 7 | 7 | | 8 | 8 | | 9 | 9 | | 10 | 10 | | 11 | 11 | | 12 | 12 | | 13 | 13 | | 14 | 14 | | 15 | 15 | | 16 | 16 | +----+------+ 16 rows in set (0.01 sec) |
但是 183 节点的数据还是 15 条。
183 节点删除一条数据
mysql> delete from ming.t1 where a=11; Query OK, 1 row affected (0.00 sec) |
两个存活节点都删除了 11 这条数据。
启动 182 节点，可以正常启动，启动后检查数据，数据与 183 一致，推测数据的 donor 节点被选择成了 183
2018-09-28T08:21:19.486736Z 2 [Note] WSREP: Check if state gap can be serviced using IST 2018-09-28T08:21:19.486832Z 2 [Note] WSREP: IST receiver addr using tcp://172.31.217.182:4568 2018-09-28T08:21:19.487026Z 2 [Note] WSREP: Prepared IST receiver, listening at: tcp://172.31.217.182:4568 2018-09-28T08:21:19.487050Z 2 [Note] WSREP: State gap can be likely serviced using IST. SST request though present would be void. 2018-09-28T08:21:19.487984Z 0 [Note] WSREP: may fallback to sst. ist_seqno [51] < safe_ist_seqno [52] 2018-09-28T08:21:19.488006Z 0 [Note] WSREP: Member 2.0 (bd-dev-mingshuo-182) requested state transfer from '*any*'. Selected 0.0 (bd-dev-mingshuo-183)(SYNCED) as donor. |
日志中可以看到， 182 节点被选择成为 IST receiver ，监听端口 4568 端口。选择 183 节点作为
数据的 donor 。那么数据与 183 一致也就不足为奇了。
此时数据出现了不一致，如何解决呢？
可以删除节点数据目录下文件，然后按照启动，通过 SST 全量恢复数据。
Pxc 启动时可以人为选择数据的 doner 节点。
wsrep_sst_donor 参数
关闭两个节点，加 wsrep_sst_donor 参数重新启动
mysqld_safe --defaults-file=/etc/my3307.cnf -- wsrep_sst_donor=172.31.217.89 & mysqld_safe --defaults-file=/etc/my3307.cnf -- wsrep_sst_donor=172.31.217.89 & |
分享文章：mysqlPXC集群脑裂及grastate.dat修改实验
文章起源：http://6mz.cn/article/iecpie.html