三台服务器做了mysql PXC集群

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：

2018-09-27T07:33:13.222079Z0 [Note] WSREP: Received shutdown signal. Will sleep for 10 secs beforeinitiating shutdown. pxc_maint_mode switched to SHUTDOWN

2018-09-27T07:33:23.230509Z0 [Note] WSREP: Stop replication

2018-09-27T07:33:23.230619Z0 [Note] WSREP: Closing send monitor...

2018-09-27T07:33:23.230640Z0 [Note] WSREP: Closed send monitor.

2018-09-27T07:33:23.230660Z0 [Note] WSREP: gcomm: terminating thread

2018-09-27T07:33:23.230680Z0 [Note] WSREP: gcomm: joining thread

2018-09-27T07:33:23.230827Z0 [Note] WSREP: gcomm: closing backend

2018-09-27T07:33:23.231780Z0 [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

}

)

2018-09-27T07:33:23.231867Z0 [Note] WSREP: Current view of cluster as seen by this node

view ((empty))

2018-09-27T07:33:23.232111Z0 [Note] WSREP: gcomm: closed

2018-09-27T07:33:23.232165Z0 [Note] WSREP: New COMPONENT: primary = no, bootstrap = no, my_idx = 0,memb_num = 1

2018-09-27T07:33:23.232253Z0 [Note] WSREP: Flow-control interval: [100, 100]

2018-09-27T07:33:23.232260Z0 [Note] WSREP: Trying to continue unpaused monitor

2018-09-27T07:33:23.232264Z0 [Note] WSREP: Received NON-PRIMARY.

2018-09-27T07:33:23.232268Z0 [Note] WSREP: Shifting SYNCED -> OPEN (TO: 27)

2018-09-27T07:33:23.232279Z0 [Note] WSREP: Received self-leave message.

2018-09-27T07:33:23.232285Z0 [Note] WSREP: Flow-control interval: [0, 0]

2018-09-27T07:33:23.232288Z0 [Note] WSREP: Trying to continue unpaused monitor

2018-09-27T07:33:23.232291Z0 [Note] WSREP: Received SELF-LEAVE. Closing connection.

2018-09-27T07:33:23.232295Z0 [Note] WSREP: Shifting OPEN -> CLOSED (TO: 27)

2018-09-27T07:33:23.232302Z0 [Note] WSREP: RECV thread exiting 0: Success

2018-09-27T07:33:23.232383Z2 [Note] WSREP: New cluster view: global state:c057dbc5-c16e-11e8-a1a6-825ed9079934:27, view# -1: non-Primary, number ofnodes: 1, my index: 0, protocol version 3

2018-09-27T07:33:23.232394Z2 [Note] WSREP: Setting wsrep_ready to false

2018-09-27T07:33:23.232400Z2 [Note] WSREP: wsrep_notify_cmd is not defined, skipping notification.

2018-09-27T07:33:23.232439Z2 [Note] WSREP: New cluster view: global state:c057dbc5-c16e-11e8-a1a6-825ed9079934:27, view# -1: non-Primary, number ofnodes: 0, my index: -1, protocol version 3

2018-09-27T07:33:23.232443Z2 [Note] WSREP: Setting wsrep_ready to false

2018-09-27T07:33:23.232446Z2 [Note] WSREP: wsrep_notify_cmd is not defined, skipping notification.

2018-09-27T07:33:23.232472Z2 [Note] WSREP: applier thread exiting (code:0)

2018-09-27T07:33:23.232479Z0 [Note] WSREP: recv_thread() joined.

2018-09-27T07:33:23.232502Z0 [Note] WSREP: Closing replication queue.

2018-09-27T07:33:23.232509Z0 [Note] WSREP: Closing slave action queue.

2018-09-27T07:33:23.232517Z0 [Note] Giving 2 client threads a chance to die gracefully

2018-09-27T07:33:25.232639Z0 [Note] WSREP: Waiting for active wsrep applier to exit

2018-09-27T07:33:25.232758Z1 [Note] WSREP: rollbacker thread exiting

2018-09-27T07:33:25.232994Z0 [Note] Giving 0 client threads a chance to die gracefully

2018-09-27T07:33:25.233010Z0 [Note] Shutting down slave threads

2018-09-27T07:33:25.233025Z0 [Note] Forcefully disconnecting 0 remaining clients

2018-09-27T07:33:25.233044Z0 [Note] Event Scheduler: Purging the queue. 0 events

2018-09-27T07:33:25.242788Z0 [Note] WSREP: Service thread queue flushed.

2018-09-27T07:33:25.250399Z0 [Note] WSREP: MemPool(SlaveTrxHandle): hit ratio: 0, misses: 0, in use: 0,in pool: 0

2018-09-27T07:33:25.250479Z0 [Note] WSREP: Shifting CLOSED -> DESTROYED (TO: 27)

2018-09-27T07:33:25.259428Z0 [Note] Binlog end

2018-09-27T07:33:25.261702Z0 [Note] Shutting down plugin 'ngram'

2018-09-27T07:33:25.261721Z0 [Note] Shutting down plugin 'partition'

2018-09-27T07:33:25.261726Z0 [Note] Shutting down plugin 'ARCHIVE'

2018-09-27T07:33:25.261729Z0 [Note] Shutting down plugin 'BLACKHOLE'

2018-09-27T07:33:25.261733Z0 [Note] Shutting down plugin 'INNODB_SYS_VIRTUAL'

2018-09-27T07:33:25.261736Z0 [Note] Shutting down plugin 'INNODB_CHANGED_PAGES'

2018-09-27T07:33:25.261739Z0 [Note] Shutting down plugin 'INNODB_SYS_DATAFILES'

2018-09-27T07:33:25.261741Z0 [Note] Shutting down plugin 'INNODB_SYS_TABLESPACES'

2018-09-27T07:33:25.261744Z0 [Note] Shutting down plugin 'INNODB_SYS_FOREIGN_COLS'

2018-09-27T07:33:25.261746Z0 [Note] Shutting down plugin 'INNODB_SYS_FOREIGN'

2018-09-27T07:33:25.261749Z0 [Note] Shutting down plugin 'INNODB_SYS_FIELDS'

2018-09-27T07:33:25.261751Z0 [Note] Shutting down plugin 'INNODB_SYS_COLUMNS'

2018-09-27T07:33:25.261754Z0 [Note] Shutting down plugin 'INNODB_SYS_INDEXES'

2018-09-27T07:33:25.261756Z0 [Note] Shutting down plugin 'INNODB_SYS_TABLESTATS'

2018-09-27T07:33:25.261759Z0 [Note] Shutting down plugin 'INNODB_SYS_TABLES'

2018-09-27T07:33:25.261761Z0 [Note] Shutting down plugin 'INNODB_FT_INDEX_TABLE'

2018-09-27T07:33:25.261764Z0 [Note] Shutting down plugin 'INNODB_FT_INDEX_CACHE'

2018-09-27T07:33:25.261766Z0 [Note] Shutting down plugin 'INNODB_FT_CONFIG'

2018-09-27T07:33:25.261769Z0 [Note] Shutting down plugin 'INNODB_FT_BEING_DELETED'

2018-09-27T07:33:25.261771Z0 [Note] Shutting down plugin 'INNODB_FT_DELETED'

2018-09-27T07:33:25.261774Z0 [Note] Shutting down plugin 'INNODB_FT_DEFAULT_STOPWORD'

2018-09-27T07:33:25.261776Z0 [Note] Shutting down plugin 'INNODB_METRICS'

2018-09-27T07:33:25.261778Z0 [Note] Shutting down plugin 'INNODB_TEMP_TABLE_INFO'

2018-09-27T07:33:25.261781Z0 [Note] Shutting down plugin 'INNODB_BUFFER_POOL_STATS'

2018-09-27T07:33:25.261783Z0 [Note] Shutting down plugin 'INNODB_BUFFER_PAGE_LRU'

2018-09-27T07:33:25.261785Z0 [Note] Shutting down plugin 'INNODB_BUFFER_PAGE'

2018-09-27T07:33:25.261788Z0 [Note] Shutting down plugin 'INNODB_CMP_PER_INDEX_RESET'

2018-09-27T07:33:25.261790Z0 [Note] Shutting down plugin 'INNODB_CMP_PER_INDEX'

2018-09-27T07:33:25.261793Z0 [Note] Shutting down plugin 'INNODB_CMPMEM_RESET'

2018-09-27T07:33:25.261795Z0 [Note] Shutting down plugin 'INNODB_CMPMEM'

2018-09-27T07:33:25.261797Z0 [Note] Shutting down plugin 'INNODB_CMP_RESET'

2018-09-27T07:33:25.261800Z0 [Note] Shutting down plugin 'INNODB_CMP'

2018-09-27T07:33:25.261802Z0 [Note] Shutting down plugin 'INNODB_LOCK_WAITS'

2018-09-27T07:33:25.261805Z0 [Note] Shutting down plugin 'INNODB_LOCKS'

2018-09-27T07:33:25.261807Z0 [Note] Shutting down plugin 'INNODB_TRX'

2018-09-27T07:33:25.261809Z0 [Note] Shutting down plugin 'XTRADB_ZIP_DICT_COLS'

2018-09-27T07:33:25.261812Z0 [Note] Shutting down plugin 'XTRADB_ZIP_DICT'

2018-09-27T07:33:25.261814Z0 [Note] Shutting down plugin 'XTRADB_RSEG'

2018-09-27T07:33:25.261817Z0 [Note] Shutting down plugin 'XTRADB_INTERNAL_HASH_TABLES'

2018-09-27T07:33:25.261819Z0 [Note] Shutting down plugin 'XTRADB_READ_VIEW'

2018-09-27T07:33:25.261822Z0 [Note] Shutting down plugin 'InnoDB'

2018-09-27T07:33:25.261857Z0 [Note] InnoDB: FTS optimize thread exiting.

2018-09-27T07:33:25.262097Z0 [Note] InnoDB: Starting shutdown...

2018-09-27T07:33:25.362428Z0 [Note] InnoDB: Dumping buffer pool(s) to/u01/mysql/3307/data/ib_buffer_pool

2018-09-27T07:33:25.363022Z0 [Note] InnoDB: Buffer pool(s) dump completed at 180927 15:33:25

2018-09-27T07:33:25.562786Z0 [Note] InnoDB: Waiting for page_cleaner to finish flushing of buffer pool

2018-09-27T07:33:26.571050Z0 [Note] InnoDB: Shutdown completed; log sequence number 2569669

2018-09-27T07:33:26.574169Z0 [Note] InnoDB: Removed temporary tablespace data file: "ibtmp1"

2018-09-27T07:33:26.574193Z0 [Note] Shutting down plugin 'MyISAM'

2018-09-27T07:33:26.574210Z0 [Note] Shutting down plugin 'MRG_MYISAM'

2018-09-27T07:33:26.574222Z0 [Note] Shutting down plugin 'CSV'

2018-09-27T07:33:26.574233Z0 [Note] Shutting down plugin 'MEMORY'

2018-09-27T07:33:26.574254Z0 [Note] Shutting down plugin 'PERFORMANCE_SCHEMA'

2018-09-27T07:33:26.574287Z0 [Note] Shutting down plugin 'sha256_password'

2018-09-27T07:33:26.574296Z0 [Note] Shutting down plugin 'mysql_native_password'

2018-09-27T07:33:26.574304Z0 [Note] Shutting down plugin 'wsrep'

2018-09-27T07:33:26.574480Z0 [Note] Shutting down plugin 'binlog'

正常节点log

2018-09-27T07:33:22.505216Z0 [Note] WSREP: declaring c05737fd at tcp://172.31.217.89:4567 stable

2018-09-27T07:33:22.505345Z0 [Note] WSREP: forgetting 12f1e199 (tcp://172.31.217.183:4567)

2018-09-27T07:33:22.511586Z0 [Note] WSREP: Node 2331d3d7 state primary

2018-09-27T07:33:22.512245Z0 [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

}

)

2018-09-27T07:33:22.512303Z0 [Note] WSREP: Save the discovered primary-component to disk

2018-09-27T07:33:22.512547Z0 [Note] WSREP: New COMPONENT: primary = yes, bootstrap = no, my_idx = 0,memb_num = 2

2018-09-27T07:33:22.513157Z0 [Note] WSREP: forgetting 12f1e199 (tcp://172.31.217.183:4567)

2018-09-27T07:33:22.513241Z0 [Note] WSREP: STATE_EXCHANGE: sent state UUID:9ccf0351-c227-11e8-ae6a-d3cac5b411a7

2018-09-27T07:33:22.514096Z0 [Note] WSREP: STATE EXCHANGE: sent state msg:9ccf0351-c227-11e8-ae6a-d3cac5b411a7

2018-09-27T07:33:22.514647Z0 [Note] WSREP: STATE EXCHANGE: got state msg: 9ccf0351-c227-11e8-ae6a-d3cac5b411a7from 0 (bd-dev-mingshuo-182)

2018-09-27T07:33:22.514661Z0 [Note] WSREP: STATE EXCHANGE: got state msg:9ccf0351-c227-11e8-ae6a-d3cac5b411a7 from 1 (bd-dev-vertica-89)

2018-09-27T07:33:22.514669Z0 [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

2018-09-27T07:33:22.514675Z0 [Note] WSREP: Flow-control interval: [141, 141]

2018-09-27T07:33:22.514679Z0 [Note] WSREP: Trying to continue unpaused monitor

2018-09-27T07:33:22.514707Z2 [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

2018-09-27T07:33:22.514713Z2 [Note] WSREP: Setting wsrep_ready to true

2018-09-27T07:33:22.514719Z2 [Note] WSREP: wsrep_notify_cmd is not defined, skipping notification.

2018-09-27T07:33:22.514727Z2 [Note] WSREP: REPL Protocols: 8 (3, 2)

2018-09-27T07:33:22.514747Z2 [Note] WSREP: Assign initial position for certification: 27, protocolversion: 3

2018-09-27T07:33:22.514830Z0 [Note] WSREP: Service thread queue flushed.

2018-09-27T07:33:27.691129Z0 [Note] WSREP: cleaning up 12f1e199(tcp://172.31.217.183:4567)

182节点插入数据

mysql> insertinto t1 values (4,4);

Query OK, 1 rowaffected (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。

2018-09-27T08:05:50.785769Z0 [Note] WSREP: Signalling provider to continue on SST completion.

2018-09-27T08:05:50.785808Z0 [Note] WSREP: Initialized wsrep sidno 2

2018-09-27T08:05:50.785833Z0 [Note] WSREP: SST received: c057dbc5-c16e-11e8-a1a6-825ed9079934:27

2018-09-27T08:05:50.785872Z2 [Note] WSREP: Receiving IST: 1 writesets, seqnos 27-28

2018-09-27T08:05:50.785985Z0 [Note]

2018-09-27T08:05:50.785985Z0 [Note] WSREP: Receiving IST... 0.0%(0/1 events) complete.

2018-09-27T08:05:50.877679Z0 [Note] WSREP: Receiving IST...100.0% (1/1 events) complete.

2018-09-27T08:05:50.877904Z2 [Note] WSREP: IST received: c057dbc5-c16e-11e8-a1a6-825ed9079934:28

2018-09-27T08:05:50.878589Z0 [Note] WSREP: 1.0 (bd-dev-mingshuo-183): State transfer from 0.0(bd-dev-mingshuo-182) complete.

2018-09-27T08:05:50.878603Z0 [Note] WSREP: SST leaving flow control

2018-09-27T08:05:50.878608Z0 [Note] WSREP: Shifting JOINER -> JOINED (TO: 28)

2018-09-27T08:05:50.879059Z0 [Note] WSREP: Member 1.0 (bd-dev-mingshuo-183) synced with group.

2018-09-27T08:05:50.879072Z0 [Note] WSREP: Shifting JOINED -> SYNCED (TO: 28)

2018-09-27T08:05:50.879101Z2 [Note] WSREP: Synchronized with group, ready for connections

2018-09-27T08:05:50.879115Z2 [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;

ERROR1047 (08S01): WSREP has not yet prepared node for application use

mysql> insertinto ming.t1 values(10,10);

ERROR1047 (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> showstatus 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> showvariables 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

2018-09-28T02:57:37.209095Z0 [Note] WSREP: (725136c0, 'tcp://0.0.0.0:4567') reconnecting to 2331d3d7(tcp://172.31.217.182:4567), attempt 960

2018-09-28T02:58:12.714612Z0 [Note] WSREP: (725136c0, 'tcp://0.0.0.0:4567') reconnecting to 252da778(tcp://172.31.217.183:4567), attempt 900

2018-09-28T02:58:22.216078Z0 [Note] WSREP: (725136c0, 'tcp://0.0.0.0:4567') reconnecting to 2331d3d7(tcp://172.31.217.182:4567), attempt 990

2018-09-28T02:58:57.721850Z0 [Note] WSREP: (725136c0, 'tcp://0.0.0.0:4567') reconnecting to 252da778(tcp://172.31.217.183:4567), attempt 930

2018-09-28T02:59:07.223430Z0 [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

2018-09-28T03:04:57.003914Z0 [Note] WSREP: (725136c0, 'tcp://0.0.0.0:4567') connection established to252da778 tcp://172.31.217.183:4567

2018-09-28T03:05:03.215507Z0 [Note] WSREP: declaring 252da778 at tcp://172.31.217.183:4567 stable

2018-09-28T03:05:03.216346Z0 [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

}

)

2018-09-28T03:05:03.216630Z0 [Note] WSREP: New COMPONENT: primary = no, bootstrap = no, my_idx = 1,memb_num = 2

2018-09-28T03:05:03.216710Z0 [Note] WSREP: Flow-control interval: [141, 141]

2018-09-28T03:05:03.216718Z0 [Note] WSREP: Trying to continue unpaused monitor

2018-09-28T03:05:03.216723Z0 [Note] WSREP: Received NON-PRIMARY.

2018-09-28T03:05:03.216794Z1 [Note] WSREP: New cluster view: global state:c057dbc5-c16e-11e8-a1a6-825ed9079934:33, view# -1: non-Primary, number ofnodes: 2, my index: 1, protocol version 3

2018-09-28T03:05:03.216822Z1 [Note] WSREP: Setting wsrep_ready to false

2018-09-28T03:05:03.216833Z1 [Note] WSREP: wsrep_notify_cmd is not defined, skipping notification.

2018-09-28T03:05:04.277523Z0 [Note] WSREP: (725136c0, 'tcp://0.0.0.0:4567') connection established to2331d3d7 tcp://172.31.217.182:4567

2018-09-28T03:05:04.279018Z0 [Note] WSREP: (725136c0, 'tcp://0.0.0.0:4567') connection established to2331d3d7 tcp://172.31.217.182:4567

2018-09-28T03:05:04.776965Z0 [Note] WSREP: declaring 2331d3d7 at tcp://172.31.217.182:4567 stable

2018-09-28T03:05:04.777019Z0 [Note] WSREP: declaring 252da778 at tcp://172.31.217.183:4567 stable

2018-09-28T03:05:04.777487Z0 [Note] WSREP: re-bootstrapping prim from partitioned components

2018-09-28T03:05:04.778262Z0 [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 {

}

)

2018-09-28T03:05:04.778307Z0 [Note] WSREP: Save the discovered primary-component to disk

2018-09-28T03:05:04.778588Z0 [Note] WSREP: New COMPONENT: primary = yes, bootstrap = no, my_idx = 2,memb_num = 3

2018-09-28T03:05:04.778629Z0 [Note] WSREP: STATE EXCHANGE: Waiting for state UUID.

2018-09-28T03:05:05.277931Z0 [Note] WSREP: STATE EXCHANGE: sent state msg:838d4806-c2cb-11e8-8bb1-eeeae1741165

2018-09-28T03:05:05.278435Z0 [Note] WSREP: STATE EXCHANGE: got state msg:838d4806-c2cb-11e8-8bb1-eeeae1741165 from 0 (bd-dev-mingshuo-182)

2018-09-28T03:05:05.278463Z0 [Note] WSREP: STATE EXCHANGE: got state msg:838d4806-c2cb-11e8-8bb1-eeeae1741165 from 1 (bd-dev-mingshuo-183)

2018-09-28T03:05:05.278470Z0 [Note] WSREP: STATE EXCHANGE: got state msg:838d4806-c2cb-11e8-8bb1-eeeae1741165 from 2 (bd-dev-vertica-89)

2018-09-28T03:05:05.278490Z0 [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

Primseqno : -1

First seqno : -1

Lastseqno : 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

Primseqno : -1

First seqno : -1

Lastseqno : 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

Primseqno : 26

First seqno : 33

Lastseqno : 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'

2018-09-28T03:05:05.278511Z0 [Note] WSREP: Partial re-merge of primary19faf204-c2c7-11e8-b642-52dd65ccae43 found: 1 of 2.

2018-09-28T03:05:05.278520Z0 [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

2018-09-28T03:05:05.278540Z0 [Note] WSREP: Flow-control interval: [173, 173]

2018-09-28T03:05:05.278544Z0 [Note] WSREP: Trying to continue unpaused monitor

2018-09-28T03:05:05.278548Z0 [Note] WSREP: Restored state OPEN -> SYNCED (33)

2018-09-28T03:05:05.278593Z1 [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

2018-09-28T03:05:05.278612Z1 [Note] WSREP: Setting wsrep_ready to true

2018-09-28T03:05:05.278621Z1 [Note] WSREP: wsrep_notify_cmd is not defined, skipping notification.

2018-09-28T03:05:05.278661Z1 [Note] WSREP: REPL Protocols: 8 (3, 2)

2018-09-28T03:05:05.278679Z1 [Note] WSREP: Assign initial position for certification: 33, protocolversion: 3

2018-09-28T03:05:05.278752Z0 [Note] WSREP: Service thread queue flushed.

2018-09-28T03:05:05.278828Z1 [Note] WSREP: Synchronized with group, ready for connections

2018-09-28T03:05:05.278849Z1 [Note] WSREP: Setting wsrep_ready to true

2018-09-28T03:05:05.278863Z1 [Note] WSREP: wsrep_notify_cmd is not defined, skipping notification.

2018-09-28T03:05:05.279134Z0 [Note] WSREP: Member 0.0 (bd-dev-mingshuo-182) synced with group.

2018-09-28T03:05:05.279179Z0 [Note] WSREP: Member 1.0 (bd-dev-mingshuo-183) synced with group.

2018-09-28T03:05:07.290328Z0 [Note] WSREP: (725136c0, 'tcp://0.0.0.0:4567') turning message relayrequesting off

出现脑裂后解决方法：

SET GLOBALwsrep_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#moregrastate.dat

# GALERA savedstate

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#moregrastate.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#moregrastate.dat

# GALERA savedstate

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> insertinto ming.t1 values (16,16);

Query OK, 1 rowaffected (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#moregrastate.dat

# GALERA savedstate

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

2018-09-28T08:13:33.156722Z0 [ERROR] WSREP: gcs/src/gcs_group.cpp:group_post_state_exchange():322:Reversing history: 52 -> 51, this member has applied 1 more events thanthe 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> deletefrom ming.t1 where a=11;

Query OK, 1 rowaffected (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

2018-09-28T08:21:19.486736Z2 [Note] WSREP: Check if state gap can be serviced using IST

2018-09-28T08:21:19.486832Z2 [Note] WSREP: IST receiver addr using tcp://172.31.217.182:4568

2018-09-28T08:21:19.487026Z2 [Note] WSREP: Prepared IST receiver, listening at:tcp://172.31.217.182:4568

2018-09-28T08:21:19.487050Z2 [Note] WSREP: State gap can be likely serviced using IST. SST requestthough present would be void.

2018-09-28T08:21:19.487984Z0 [Note] WSREP: may fallback to sst. ist_seqno [51] < safe_ist_seqno [52]

2018-09-28T08:21:19.488006Z0 [Note] WSREP: Member 2.0 (bd-dev-mingshuo-182) requested state transferfrom '*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&