梳理下MySQL崩潰恢復過程

基於MySQL5.7版本，5.7版本在恢復過程做了優化，本文描述不考慮之前版本。

1 初始化崩潰恢復

    數據庫關閉只有2種情況，正常關閉，非正常關閉（包括數據庫實例crash及服務器crash）。

    正常關閉情況，所有buffer pool里邊的臟頁都會都會刷新一遍到磁盤，同時記錄最新LSN到ibdata文件的第一個page中。而非正常關閉來不及做這些操作，也就是沒及時把臟數據flush到磁盤，也沒有記錄最新LSN到ibdata file。

    當我們重啟數據庫實例的時候，數據庫做2個階段性操作：redo log處理，undo log及binlog 處理。

1.1 redo log處理

• 打開系統表空間ibdata，讀取第一個page中的LSN，若第一個頁損壞，則依次往后面的page讀，知道有個完整的page能夠提供LSN，這個LSN當作上次shutdown時的checkpoint點，后續恢復，從這個LSN點開始恢復

• 進入redo log文件，讀取第一個redo log文件頭的checkpoint LSN， 並根據該LSN定位到redo日志文件中對應的位置，從該checkpoint點開始掃描，進行3次redo log文件掃描：

• • 第一次，找 MLOG_CHECKPOINT日志

• • • 如果是正常關閉，這個日志是不做記錄的，也就是掃描的過程中不回找到對應的MLOG_CHECKPOINT日志，不會進行接下來的兩次掃描，因為屬於正常關閉數據庫服務，不需要考慮奔潰恢復情況；

• • • 如果是非正常關閉，則會查找到 MLOG_CHECKPOINT （如果是多個，則說明redo文件已損壞，恢復報錯），獲取MLOG_FILE_NAME中指定后續需要恢復的ibd文件；

• • 第二次，從redo log讀到的LSN，找到checkpoint點開始重復掃描存儲日志對象

• • • 根據MLOG_CHECKPOINT日志，讀取對應LSN之后的日志解析到hash表中，如果剩下的日志解析結束后還沒有填滿hash表格，則不需要進行第三次掃描；

• • • 進行到這里，則說明數據庫是非正常關閉，會在errorlog中提示：Database was not shutdown normally!詳見下圖。

• • • 

• • 第三次，若第二次掃描hash表空間不足，則發起第三次掃描，清空hash表空間，重新從新的checkpoint點開始掃描

• • • 如果第二次掃掃描就把hash表填滿，那么會先把hash表里邊的記錄重做到buffer pool中的數據頁，然后再來加載redo log 記錄到被清空的hash表中，hash表滿后立即執行恢復操作，知道所有需要redo 的redo log 被應用結束。

    恢復的過程中，注意兩個點：打開ibd文件形式，讀取數據庫到buffer pool的改進。

    根據hash表中的相應信息讀取數據頁， 讀數據頁的時候，5.7之前版本采用把所有表空間都打開，所有表格僅執行ReadOnly，5.7版本做了優化，新增了 MLOG_FILE_NAME 記錄在checkpoint之后，所有被修改過的信息，根據這些信息，在恢復過程中，只需要打開相應的ibd文件即可，不涉及恢復的表格支持正常DML跟DDL操作，涉及恢復的表格則僅執行ReadOnly功能。

    當把數據頁讀取到buffer pool中，以往版本是只讀取對應的 單個頁面，而現在的是直接讀取與該頁面相鄰的32個data page 也一起加載的buffer pool，因為一個數據頁的修改，可能周圍的頁面也被修改，一次性讀取，可以避免后面根據hash表中再重新讀取其相鄰的頁面。

1.2 undo log及binlog 處理

      上一階段中，把redo log中的操作都apply到數據頁中，但是對於prepare狀態的事務卻還沒有進行回滾處理，這個階段則是針對prepare狀態的事務進行處理，需要使用到binlog和undo log。

• 根據最后一個binlog文件，為啥不是所有binlog文件呢？因為每一個binlog文件切換的時候，都會確保當前binlog文件的所有操作已落盤，所以只需要考慮最后一個binlog文件。跟進最后一個binlog文件，獲取所有可能沒有提交事務的xid列表；

• 根據undo log中的 insert_undo_list，upddate_undo_list事務鏈，構建undo_list，在根據undo_list構建未提交事務鏈表;

• 從未提交事務鏈表中，提取出xid，凡是存在於xid列表中的，則需要提交，不存在的，則回滾。

      整個恢復過程，可以參考下來自 www.sysdb.cn 網站作者  boyce 畫的說明圖，圖片版權屬於該作者，本處僅為引用分享給大家，作圖很詳細，一言不合開源碼分析！遺憾的是，作者只寫了2篇博文就停止更新了，心疼默哀十分鍾.....

2 innodb關閉恢復涉及參數

2.1 關閉參數：innodb_fast_shutdown

    數據庫關閉的時候，innodb需要完成所有full purge何insert buffer操作，這需要一些時間，甚至幾個小時完成

• 0
  • 關閉時，需要完成所有full purge，insert buffer， flush dirty pages操作
  • 做 innodb plugin升級時，會設置為0，再正常關閉數據庫。
• 1
  • 默認值，表示不需要完成full purge和insert buffer操作，但是緩沖池的臟數據需要刷新到磁盤
• 2
  • 既不完成full purge和insert buffer操作，也不把緩沖池的臟數據刷新到磁盤，而是將日志寫入日志文件，這樣不回有數據丟失，但是下次mysql啟動的時候，需要根據日志文件執行恢復操作

2.2 恢復參數：innodb_force_recovery

    這里注意，數字越小，則忽略檢查的內容越少，每個大的數字都包含了前面小數字忽略檢查的內容。當參數設置大於0后，可以對表格進行DML操作，但是DDL操作時不允許的。

    當innodb_force_recovery 值為1-3時，僅允許SELECT TABLE ，DROP or CREATE tables；innodb_force_recovery 值為>=4時，5.7.17之前版本支持DROP TABLE，5.7.18后版本不支持。

    參數說明如下圖：

3 測試情況

    測試內容：

    begin;insert into orders select * from orders.orders limit 100000;

    插入結束后不提交事務，執行 kill mysql進程

    tailf error.log查看

3.1 默認配置測試：innodb_fast_shutdown=1，innodb_force_recovery=0

   配置說明：異常停止DB服務時，不需要完成 full purge 和 insert buffer 操作，但是緩沖池的臟數據需要刷新到磁盤；服務啟動的時候，做所有需要的檢查跟事務操作。

3.1.1 操作過程

#1 tailf error.log查看mysql錯誤日志,動態滾動查看
mysql> show global variables like 'log_error';
tailf /data/mysql/mysql3310/data/error.log

#2 測試庫中開啟事務，insert 10w行記錄，不提交事務
mysql> begin;insert into orders select * from orders.orders limit 100000;
Query OK, 0 rows affected (0.00 sec)

Query OK, 100000 rows affected (37.55 sec)
Records: 100000  Duplicates: 0  Warnings: 0

#3 查找mysql進程號，殺進程
[root@localhost opt]# ps axu | grep mysql
[root@localhost opt]# kill -9 mysql的進程號

#4 啟動mysql服務
mysqld --defaults-file=/data/mysql/mysql3310/mysql3310.cnf &

#5 到tailf error.log窗口查看錯誤

3.1.2 error log解析

啟動mysqld進程

2017-03-16T16:28:13.812074Z 0 [Note] mysqld (mysqld 5.7.14-log) starting as process 28091 ...

 

讀取ibdata的LSN

讀取redo文件的checkpoint，從該點的LSN開始查找MLOG_CHECKPOINT

發現MLOG_CHECKPOINT日志，提示：Database was not shutdown normally !  

2017-03-16T16:28:16.572276Z 0 [Note] InnoDB: Log scan progressed past the checkpoint lsn 898634772645

2017-03-16T16:28:16.572418Z 0 [Note] InnoDB: Doing recovery: scanned up to log sequence number 898634772654

2017-03-16T16:28:16.693771Z 0 [Note] InnoDB: Doing recovery: scanned up to log sequence number 898634772654

2017-03-16T16:28:16.693880Z 0 [Note] InnoDB: Database was not shutdown normally!

2017-03-16T16:28:16.693913Z 0 [Note] InnoDB: Starting crash recovery.

 

檢測需要重做跟回滾的事務

由於沒有需要redo的事務，但是有undo事務，開始undo操作

加載redo log中的96個系統回滾段 rollbak segment，32個不需要redo 的臨時表空間的回滾段

2017-03-16T16:28:17.360953Z 0 [Note] InnoDB: 1 transaction(s) which must be rolled back or cleaned up in total 100000 row operations to undo

 

獲取binlog回滾xid列表，結合undo log做回滾操作

2017-03-16T16:28:17.361122Z 0 [Note] InnoDB: Trx id counter is 5716480

2017-03-16T16:28:17.625441Z 0 [Note] InnoDB: Last MySQL binlog file position 0 37774, file name bin_log.000016

2017-03-16T16:28:17.787684Z 0 [Note] InnoDB: Starting in background the rollback of uncommitted transactions

2017-03-16T16:28:17.788022Z 0 [Note] InnoDB: Rolling back trx with id 5715975, 100000 rows to undo

 

啟動innodb，加載buffer pool

2017-03-16T16:28:17.838323Z 0 [Note] InnoDB: Waiting for purge to start

2017-03-16T16:28:18.015792Z 0 [Note] InnoDB: 5.7.14 started; log sequence number 898634772654

 

mysqld正常運行提供用戶使用

2017-03-16T16:28:18.772035Z 0 [Note] mysqld: ready for connections.

Version: '5.7.14-log'  socket: '/tmp/mysql3310.sock'  port: 3310  MySQL Community Server (GPL)

 

這個時候，recovery可能還沒結束，比如這個例子中，recovery就還在進行，其涉及的表僅提供ReadOnly操作，其他表格可正常操作 ，慢慢恢復，最后undo回滾事務完成，Rollback of non-prepared transactions completed

 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 1002017-03-16T16:29:24.353056Z 0 [Note] InnoDB: Rollback of trx with id 5715975 completed

2017-03-16T16:29:24.353267Z 0 [Note] InnoDB: Rollback of non-prepared transactions completed

 

整個過程16:28:13啟動，16:28:18 DB提供部分服務，16:29:24全庫恢復，db部分提供服務耗時5s，全庫提供服務耗時71s 。

2017-03-16T16:28:13.812074Z 0 [Note] mysqld (mysqld 5.7.14-log) starting as process 28091 ...
2017-03-16T16:28:14.250956Z 0 [Note] InnoDB: PUNCH HOLE support not available
2017-03-16T16:28:14.251212Z 0 [Note] InnoDB: Mutexes and rw_locks use GCC atomic builtins
2017-03-16T16:28:14.251311Z 0 [Note] InnoDB: Uses event mutexes
2017-03-16T16:28:14.251359Z 0 [Note] InnoDB: GCC builtin __sync_synchronize() is used for memory barrier
2017-03-16T16:28:14.251401Z 0 [Note] InnoDB: Compressed tables use zlib 1.2.3
2017-03-16T16:28:14.251311Z 0 [Note] InnoDB: Uses event mutexes
2017-03-16T16:28:14.251359Z 0 [Note] InnoDB: GCC builtin __sync_synchronize() is used for memory barrier
2017-03-16T16:28:14.251401Z 0 [Note] InnoDB: Compressed tables use zlib 1.2.3
2017-03-16T16:28:14.251440Z 0 [Note] InnoDB: Using Linux native AIO
2017-03-16T16:28:14.252570Z 0 [Note] InnoDB: Number of pools: 1
2017-03-16T16:28:14.253079Z 0 [Note] InnoDB: Not using CPU crc32 instructions
2017-03-16T16:28:14.258163Z 0 [Note] InnoDB: Initializing buffer pool, total size = 9G, instances = 8, chunk size = 128M
2017-03-16T16:28:16.101862Z 0 [Note] InnoDB: Completed initialization of buffer pool
2017-03-16T16:28:16.406890Z 0 [Note] InnoDB: If the mysqld execution user is authorized, page cleaner thread priority can be changed. See the man page of setpriority().
2017-03-16T16:28:16.447145Z 0 [Note] InnoDB: Highest supported file format is Barracuda.
2017-03-16T16:28:16.572276Z 0 [Note] InnoDB: Log scan progressed past the checkpoint lsn 898634772645
2017-03-16T16:28:16.572418Z 0 [Note] InnoDB: Doing recovery: scanned up to log sequence number 898634772654
2017-03-16T16:28:16.693771Z 0 [Note] InnoDB: Doing recovery: scanned up to log sequence number 898634772654
2017-03-16T16:28:16.693880Z 0 [Note] InnoDB: Database was not shutdown normally!
2017-03-16T16:28:16.693913Z 0 [Note] InnoDB: Starting crash recovery.
2017-03-16T16:28:17.360953Z 0 [Note] InnoDB: 1 transaction(s) which must be rolled back or cleaned up in total 100000 row operations to undo
2017-03-16T16:28:17.361122Z 0 [Note] InnoDB: Trx id counter is 5716480
2017-03-16T16:28:17.625441Z 0 [Note] InnoDB: Last MySQL binlog file position 0 37774, file name bin_log.000016
2017-03-16T16:28:17.787684Z 0 [Note] InnoDB: Starting in background the rollback of uncommitted transactions
2017-03-16T16:28:17.788022Z 0 [Note] InnoDB: Rolling back trx with id 5715975, 100000 rows to undo

InnoDB: Progress in percents: 12017-03-16T16:28:17.788272Z 0 [Note] InnoDB: Removed temporary tablespace data file: "ibtmp1"
2017-03-16T16:28:17.788603Z 0 [Note] InnoDB: Creating shared tablespace for temporary tables
2017-03-16T16:28:17.788603Z 0 [Note] InnoDB: Creating shared tablespace for temporary tables
2017-03-16T16:28:17.788937Z 0 [Note] InnoDB: Setting file './ibtmp1' size to 12 MB. Physically writing the file full; Please wait ...
2017-03-16T16:28:17.831078Z 0 [Note] InnoDB: File './ibtmp1' size is now 12 MB.
2017-03-16T16:28:17.834673Z 0 [Note] InnoDB: 96 redo rollback segment(s) found. 96 redo rollback segment(s) are active.
2017-03-16T16:28:17.834822Z 0 [Note] InnoDB: 32 non-redo rollback segment(s) are active.
2017-03-16T16:28:17.838323Z 0 [Note] InnoDB: Waiting for purge to start
2017-03-16T16:28:18.015792Z 0 [Note] InnoDB: 5.7.14 started; log sequence number 898634772654
2017-03-16T16:28:18.018091Z 0 [Note] InnoDB: Loading buffer pool(s) from /data/mysql/mysql3310/data/ib_buffer_pool
2017-03-16T16:28:18.018464Z 0 [Note] Plugin 'FEDERATED' is disabled.
2017-03-16T16:28:18.119301Z 0 [Note] Recovering after a crash using /data/mysql/mysql3310/logs/bin_log
2017-03-16T16:28:18.119442Z 0 [Note] Starting crash recovery...
2017-03-16T16:28:18.119442Z 0 [Note] Starting crash recovery...
2017-03-16T16:28:18.119624Z 0 [Note] Crash recovery finished.
2017-03-16T16:28:18.203723Z 0 [Warning] Failed to set up SSL because of the following SSL library error: SSL context is not usable without certificate and private key
2017-03-16T16:28:18.203835Z 0 [Note] Server hostname (bind-address): '*'; port: 3310
2017-03-16T16:28:18.204188Z 0 [Note] IPv6 is available.
2017-03-16T16:28:18.204339Z 0 [Note]   - '::' resolves to '::';
2017-03-16T16:28:18.204667Z 0 [Note] Server socket created on IP: '::'.
2017-03-16T16:28:18.377789Z 0 [Warning] 'user' entry 'mysql.sys@localhost' ignored in --skip-name-resolve mode.
2017-03-16T16:28:18.378054Z 0 [Warning] 'db' entry 'sys mysql.sys@localhost' ignored in --skip-name-resolve mode.
2017-03-16T16:28:18.378170Z 0 [Warning] 'proxies_priv' entry '@ root@localhost' ignored in --skip-name-resolve mode.
2017-03-16T16:28:18.392680Z 0 [Warning] 'tables_priv' entry 'sys_config mysql.sys@localhost' ignored in --skip-name-resolve mode.
2017-03-16T16:28:18.771122Z 0 [Note] Event Scheduler: Loaded 0 events
2017-03-16T16:28:18.772035Z 0 [Note] mysqld: ready for connections.
Version: '5.7.14-log'  socket: '/tmp/mysql3310.sock'  port: 3310  MySQL Community Server (GPL)
2017-03-16T16:28:21.404135Z 0 [Note] InnoDB: Buffer pool(s) load completed at 170317  0:28:21
 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 1002017-03-16T16:29:24.353056Z 0 [Note] InnoDB: Rollback of trx with id 5715975 completed
2017-03-16T16:29:24.353267Z 0 [Note] InnoDB: Rollback of non-prepared transactions completed

3.2 innodb_fast_shutdown=1，innodb_force_recovery=3

    配置說明：異常停止DB服務時，不需要完成full purge和insert buffer操作，但是緩沖池的臟數據需要刷新到磁盤；服務啟動的時候，忽略檢查corrupt頁，也阻止主線程的運行，並且不執行事務回滾操作。

    當innodb_force_recovery 值為1-3時，允許SELECT TABLE ，DROP or CREATE tables；innodb_force_recovery 值為>=4時，5.7.17之前版本支持DROP TABLE，5.7.18后版本不支持。

    這個配置適用於這種情況：有某些大表執行alter或者大量數據修改操作生成大量undo宕機，那么這個情況下，可以進入cnf配置文件修改innodb_force_recovery=3，然后把再做undo操作的表格數據導出來或者利用備份文件的數據，再drop掉這個表格，進入cnf配置文件，修改參數為0，重啟DB服務，會發現，rollback過程非常快執行結束，因為找不到對應表格，所以rollback非常快。

 

    Tips: 如何查找在undo操作的表格？

    數據庫crash后，第一個正常啟動恢復連接，連進數據庫里邊，操作所有表格，不支持insert update 的表格都屬於正在undo的表。

 

3.2.1 操作過程

#1 tailf error.log查看mysql錯誤日志,動態滾動查看
mysql> show global variables like 'log_error';
tailf /data/mysql/mysql3310/data/error.log

#2 測試庫中開啟事務，insert 10w行記錄，不提交事務
mysql> begin;insert into orders select * from orders.orders limit 100000;
Query OK, 0 rows affected (0.00 sec)

Query OK, 100000 rows affected (37.55 sec)
Records: 100000  Duplicates: 0  Warnings: 0

#3 查找mysql進程號，殺進程
[root@localhost opt]# ps axu | grep mysql
[root@localhost opt]# kill -9 mysql的進程號

#4 在cnf文件中指定innodb_force_recovery=3,啟動服務，檢查是否修改成功
[root@localhost ~]# vim /data/mysql/mysql3310.cnf
#添加innodb_force_recovery參數設置
[mysqld]
innodb_force_recovery=3

#5 啟動mysql服務
[root@localhost ~]# mysqld --defaults-file=/data/mysql/mysql3310.cnf &

#6 到tailf error.log窗口查看錯誤
#發現沒有進行undo操作，同時全庫僅支持select drop create，不支持其他所有操作

3.2.2 error log解析

啟動mysqld進程

2017-03-17T15:12:14.317391Z 0 [Note] mysqld (mysqld 5.7.14-log) starting as process 32094 ...

 

讀取ibdata的LSN

讀取redo文件的checkpoint，從該點的LSN開始查找MLOG_CHECKPOINT

發現MLOG_CHECKPOINT日志，提示：Database was not shutdown normally !  

2017-03-17T15:12:16.478403Z 0 [Note] InnoDB: Log scan progressed past the checkpoint lsn 898935661420

2017-03-17T15:12:16.478579Z 0 [Note] InnoDB: Doing recovery: scanned up to log sequence number 898935661429

2017-03-17T15:12:16.532923Z 0 [Note] InnoDB: Doing recovery: scanned up to log sequence number 898935661429

2017-03-17T15:12:16.533063Z 0 [Note] InnoDB: Database was not shutdown normally!

2017-03-17T15:12:16.533105Z 0 [Note] InnoDB: Starting crash recovery.

 

檢測需要重做跟回滾的事務

由於沒有需要redo的事務，但是有undo事務，開始undo操作

加載redo log中的96個系統回滾段 rollbak segment，32個不需要redo 的臨時表空間的回滾段

2017-03-17T15:12:17.052734Z 0 [Note] InnoDB: 1 transaction(s) which must be rolled back or cleaned up in total 100000 row operations to undo

 

獲取binlog回滾xid列表，但是這里注意，沒有做 回滾操作

2017-03-17T15:12:17.053009Z 0 [Note] InnoDB: Trx id counter is 5718016

2017-03-17T15:12:17.321987Z 0 [Note] InnoDB: Last MySQL binlog file position 0 37774, file name bin_log.000016

 

啟動innodb，加載buffer pool ，警告當前 innodb_force_recovery 設置為3

2017-03-17T15:12:17.518732Z 0 [Note] InnoDB: 5.7.14 started; log sequence number 898935661429

2017-03-17T15:12:17.518789Z 0 [Note] InnoDB: !!! innodb_force_recovery is set to 3 !!!

2017-03-17T15:12:17.519305Z 0 [Note] InnoDB: Loading buffer pool(s) from /data/mysql/mysql3310/data/ib_buffer_pool

 

mysqld正常運行提供用戶使用，但是，全庫僅提供readonly

2017-03-17T15:12:18.521680Z 0 [Note] mysqld: ready for connections.

Version: '5.7.14-log'  socket: '/tmp/mysql3310.sock'  port: 3310  MySQL Community Server (GPL)

2017-03-17T15:13:03.125686Z 2 [ERROR] InnoDB: innodb_force_recovery is on. We do not allow database modifications by the user. Shut down mysqld and edit my.cnf to set innodb_force_recovery=0

 

    整個過程 15:12:14啟動，15:12:18 DB提供部分服務，耗時4s。但是由於沒有做undo操作，錯誤日志提醒 We do not allow database modifications by the user. 不支持當前用戶進行數據庫修改操作，僅支持select table，drop table，create table操作，其他所有操作都不支持。

    這種情況下啟動數據庫后，可以支持你dump出正在做undo的表格數據，然后drop掉它，如果你有備份文件恢復，那么則不需要dump，直接drop，然后修改cnf配置文件，設置 innodb_force_recovery=0 ，重啟數據庫服務，這個時候，innodb 的undo recovery非常快，因為找不到對應的表格操作，所以都是直接挑過處理。正常恢復后，趕緊把drop掉的表格從備份文件或者dump出來的sql文件中恢復回來。注意注意！！！！這種情況要權衡利弊，從時間耗時跟操作安全數據安全考慮，是undo recovery快還是你如此操作快，然后一定一定要跟你的上級或者領導報備操作的風險性，不要默默干，要報告！！！

2017-03-17T15:12:14.317391Z 0 [Note] mysqld (mysqld 5.7.14-log) starting as process 32094 ...
2017-03-17T15:12:14.430101Z 0 [Note] InnoDB: PUNCH HOLE support not available
2017-03-17T15:12:14.430586Z 0 [Note] InnoDB: Mutexes and rw_locks use GCC atomic builtins
2017-03-17T15:12:14.430732Z 0 [Note] InnoDB: Uses event mutexes
2017-03-17T15:12:14.430874Z 0 [Note] InnoDB: GCC builtin __sync_synchronize() is used for memory barrier
2017-03-17T15:12:14.430993Z 0 [Note] InnoDB: Compressed tables use zlib 1.2.3
2017-03-17T15:12:14.431084Z 0 [Note] InnoDB: Using Linux native AIO
2017-03-17T15:12:14.433127Z 0 [Note] InnoDB: Number of pools: 1
2017-03-17T15:12:14.434016Z 0 [Note] InnoDB: Not using CPU crc32 instructions
2017-03-17T15:12:14.443970Z 0 [Note] InnoDB: Initializing buffer pool, total size = 9G, instances = 8, chunk size = 128M
2017-03-17T15:12:16.158820Z 0 [Note] InnoDB: Completed initialization of buffer pool
2017-03-17T15:12:16.380118Z 0 [Note] InnoDB: If the mysqld execution user is authorized, page cleaner thread priority can be changed. See the man page of setpriority().
2017-03-17T15:12:16.409819Z 0 [Note] InnoDB: Highest supported file format is Barracuda.
2017-03-17T15:12:16.478403Z 0 [Note] InnoDB: Log scan progressed past the checkpoint lsn 898935661420
2017-03-17T15:12:16.478579Z 0 [Note] InnoDB: Doing recovery: scanned up to log sequence number 898935661429
2017-03-17T15:12:16.532923Z 0 [Note] InnoDB: Doing recovery: scanned up to log sequence number 898935661429
2017-03-17T15:12:16.533063Z 0 [Note] InnoDB: Database was not shutdown normally!
2017-03-17T15:12:16.533105Z 0 [Note] InnoDB: Starting crash recovery.
2017-03-17T15:12:17.052734Z 0 [Note] InnoDB: 1 transaction(s) which must be rolled back or cleaned up in total 100000 row operations to undo
2017-03-17T15:12:17.053009Z 0 [Note] InnoDB: Trx id counter is 5718016
2017-03-17T15:12:17.321987Z 0 [Note] InnoDB: Last MySQL binlog file position 0 37774, file name bin_log.000016
2017-03-17T15:12:17.460435Z 0 [Note] InnoDB: Removed temporary tablespace data file: "ibtmp1"
2017-03-17T15:12:17.460673Z 0 [Note] InnoDB: Creating shared tablespace for temporary tables
2017-03-17T15:12:17.461075Z 0 [Note] InnoDB: Setting file './ibtmp1' size to 12 MB. Physically writing the file full; Please wait ...
2017-03-17T15:12:17.512923Z 0 [Note] InnoDB: File './ibtmp1' size is now 12 MB.
2017-03-17T15:12:17.516337Z 0 [Note] InnoDB: 96 redo rollback segment(s) found. 96 redo rollback segment(s) are active.
2017-03-17T15:12:17.516617Z 0 [Note] InnoDB: 32 non-redo rollback segment(s) are active.
2017-03-17T15:12:17.518732Z 0 [Note] InnoDB: 5.7.14 started; log sequence number 898935661429
2017-03-17T15:12:17.518789Z 0 [Note] InnoDB: !!! innodb_force_recovery is set to 3 !!!
2017-03-17T15:12:17.519305Z 0 [Note] InnoDB: Loading buffer pool(s) from /data/mysql/mysql3310/data/ib_buffer_pool
2017-03-17T15:12:17.521217Z 0 [Note] Plugin 'FEDERATED' is disabled.
2017-03-17T15:12:17.694697Z 0 [Note] Recovering after a crash using /data/mysql/mysql3310/logs/bin_log
2017-03-17T15:12:17.694882Z 0 [Note] Starting crash recovery...
2017-03-17T15:12:17.695070Z 0 [Note] Crash recovery finished.
2017-03-17T15:12:18.197454Z 0 [Warning] Failed to set up SSL because of the following SSL library error: SSL context is not usable without certificate and private key
2017-03-17T15:12:18.197584Z 0 [Note] Server hostname (bind-address): '*'; port: 3310
2017-03-17T15:12:18.197804Z 0 [Note] IPv6 is available.
2017-03-17T15:12:18.197910Z 0 [Note]   - '::' resolves to '::';
2017-03-17T15:12:18.197981Z 0 [Note] Server socket created on IP: '::'.
2017-03-17T15:12:18.391615Z 0 [Note] InnoDB: Buffer pool(s) load completed at 170317 23:12:18
2017-03-17T15:12:18.414950Z 0 [Warning] 'user' entry 'mysql.sys@localhost' ignored in --skip-name-resolve mode.
2017-03-17T15:12:18.415223Z 0 [Warning] 'db' entry 'sys mysql.sys@localhost' ignored in --skip-name-resolve mode.
2017-03-17T15:12:18.415340Z 0 [Warning] 'proxies_priv' entry '@ root@localhost' ignored in --skip-name-resolve mode.
2017-03-17T15:12:18.446935Z 0 [Warning] 'tables_priv' entry 'sys_config mysql.sys@localhost' ignored in --skip-name-resolve mode.
2017-03-17T15:12:18.520569Z 0 [Note] Event Scheduler: Loaded 0 events
2017-03-17T15:12:18.521680Z 0 [Note] mysqld: ready for connections.
Version: '5.7.14-log'  socket: '/tmp/mysql3310.sock'  port: 3310  MySQL Community Server (GPL)
2017-03-17T15:13:03.125686Z 2 [ERROR] InnoDB: innodb_force_recovery is on. We do not allow database modifications by the user. Shut down mysqld and edit my.cnf to set innodb_force_recovery=0

 

參考文章：

https://dev.mysql.com/doc/refman/5.7/en/innodb-parameters.html

http://www.sysdb.cn/index.php/2016/01/14/innodb-recovery/?utm_source=tuicool&utm_medium=referral

http://mysql.taobao.org/monthly/2015/06/01/