歡迎關注我的《深入理解MySQL主從原理 32講 》，如下：

![image.png](https://upload-images.jianshu.io/upload_images/7398834-0ffa3bdc078cddf4.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)
如果圖片不能顯示可檢視下面連結： 
https://www.jianshu.com/p/d636215d767f

###一、stop slave流程
使用者執行緒：
```
stop_slave
 -> terminate_slave_threads 
  ->帶入引數rpl_stop_slave_timeout設定，作為等待SQL執行緒退出的超時時間。
  SQL:
  -> 設定SQL執行緒終止，mi->rli->abort_slave= 1; 這個才是決定作用的
  -> terminate_slave_thread SQL執行緒
    -> 給SQL執行緒 pthread_kill 傳送訊號SIGUSR1 但是本訊號已經捕捉 訊號處理函式函式為空
       empty_signal_handler
    -> 設定thd->awake(THD::NOT_KILLED) 
       
    -> 等待SQL執行緒退出，有用rli->run_lock和rli->stop_cond實現，期間會2納秒醒來判斷
       等待是否超過了rpl_stop_slave_timeout設定，如果超過繼續下面的流程 
  ->  if (mi->rli->flush_info(TRUE)) 強制刷replay info

IO：
  ->設定IO執行緒終止，mi->abort_slave=1;;   這個才是決定作用的
  -> terminate_slave_thread IO執行緒
     如上，只是是IO執行緒
  
  -> if (mi->flush_info(TRUE)) 強制刷maseter info
  ->  if (mi->rli->relay_log.is_open() && mi->rli->relay_log.flush_and_sync(true))強制刷relay log
```

SQL執行緒：
```
   在迴圈中判斷進行響應，如果mi->rli->abort_slave= 1不在繼續應用Event。
   
   DML回滾
   DDL等待
   
   以sql執行緒為例 在
handle_slave_sql   
   ->sql_slave_killed 判斷變數mi->rli->abort_slave是否已經被設定
```
   
 ```
rollback棧幀：
#0  trx_rollback_low (trx=0x7fffd78045f0) at /root/mysqlall/percona-server-locks-detail-5.7.22/storage/innobase/trx/trx0roll.cc:202
#1  0x0000000001bfed1c in trx_rollback_for_mysql (trx=0x7fffd78045f0) at /root/mysqlall/percona-server-locks-detail-5.7.22/storage/innobase/trx/trx0roll.cc:289
#2  0x00000000019cbb3f in innobase_rollback (hton=0x2f2c420, thd=0x7ffe78022a10, rollback_trx=true)
    at /root/mysqlall/percona-server-locks-detail-5.7.22/storage/innobase/handler/ha_innodb.cc:5126
#3  0x0000000000f808dc in ha_rollback_low (thd=0x7ffe78022a10, all=true) at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/handler.cc:2007
#4  0x0000000001879f91 in MYSQL_BIN_LOG::rollback (this=0x2e87840, thd=0x7ffe78022a10, all=true)
    at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/binlog.cc:2447
#5  0x0000000000f80b72 in ha_rollback_trans (thd=0x7ffe78022a10, all=true) at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/handler.cc:2094
#6  0x00000000016dd28d in trans_rollback (thd=0x7ffe78022a10) at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/transaction.cc:356
#7  0x00000000018da937 in Relay_log_info::cleanup_context (this=0x7ffe80024d30, thd=0x7ffe78022a10, error=true)
    at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/rpl_rli.cc:1867
#8  0x00000000018c4b7f in handle_slave_sql (arg=0x7ffe8001dbe0) at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/rpl_slave.cc:7645
#9  0x0000000001945620 in pfs_spawn_thread (arg=0x7ffe800381d0) at /root/mysqlall/percona-server-locks-detail-5.7.22/storage/perfschema/pfs.cc:2190
#10 0x00007ffff7bc6aa1 in start_thread () from /lib64/libpthread.so.0
#11 0x00007ffff6719bcd in clone () from /lib64/libc.so.6
  ```

STOP SLAVE 使用者執行緒棧幀：
```
Thread 9 (Thread 0x7fc266dce700 (LWP 26415)):
#0  0x00007fc266809d42 in pthread_cond_timedwait@@GLIBC_2.3.2 () from /lib64/libpthread.so.0
#1  0x0000000000e77514 in native_cond_timedwait (abstime=0x7fc266dcb860, mutex=0x7fc2223d3528, cond=0x7fc2223d3628) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/include/thr_cond.h:129
#2  my_cond_timedwait (abstime=0x7fc266dcb860, mp=0x7fc2223d3528, cond=0x7fc2223d3628) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/include/thr_cond.h:182
#3  inline_mysql_cond_timedwait (src_file=0x14e60d0 "/mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/sql/rpl_slave.cc", src_line=1884, abstime=0x7fc266dcb860, mutex=0x7fc2223d3528, that=0x7fc2223d3628) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/include/mysql/psi/mysql_thread.h:1229
#4  terminate_slave_thread (thd=0x7fc20b40e000, term_lock=term_lock@entry=0x7fc2223d3528, term_cond=0x7fc2223d3628, slave_running=0x7fc2223d36e4, stop_wait_timeout=stop_wait_timeout@entry=0x7fc266dcb920, need_lock_term=need_lock_term@entry=false) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/sql/rpl_slave.cc:1884
#5  0x0000000000e7a735 in terminate_slave_thread (need_lock_term=<optimized out="">, stop_wait_timeout=0x7fc266dcb920, slave_running=<optimized out="">, term_cond=<optimized out="">, term_lock=0x7fc2223d3528, thd=<optimized out="">) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/sql/rpl_slave.cc:1698
#6  terminate_slave_threads (mi=0x7fc259bde000, thread_mask=3, stop_wait_timeout=<optimized out="">, need_lock_term=<optimized out="">) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/sql/rpl_slave.cc:1698
#7  0x0000000000e87007 in stop_slave (thd=thd@entry=0x7fc20e016000, mi=0x7fc259bde000, net_report=net_report@entry=true, for_one_channel=for_one_channel@entry=false, push_temp_tables_warning=push_temp_tables_warning@entry=0x7fc266dcba10) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/sql/rpl_slave.cc:10344
#8  0x0000000000e873ac in stop_slave (thd=thd@entry=0x7fc20e016000) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/sql/rpl_slave.cc:605
#9  0x0000000000e874c9 in stop_slave_cmd (thd=thd@entry=0x7fc20e016000) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/sql/rpl_slave.cc:759
#10 0x0000000000c9b759 in mysql_execute_command (thd=thd@entry=0x7fc20e016000, first_level=first_level@entry=true) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/sql/sql_parse.cc:3618
#11 0x0000000000ca258d in mysql_parse (thd=thd@entry=0x7fc20e016000, parser_state=parser_state@entry=0x7fc266dcd020) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/sql/sql_parse.cc:5901
#12 0x0000000000ca314d in dispatch_command (thd=thd@entry=0x7fc20e016000, com_data=com_data@entry=0x7fc266dcdc80, command=COM_QUERY) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/sql/sql_parse.cc:1490
#13 0x0000000000ca4b27 in do_command (thd=thd@entry=0x7fc20e016000) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/sql/sql_parse.cc:1021
#14 0x0000000000d6a430 in handle_connection (arg=arg@entry=0x7fc221fbbeb0) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/sql/conn_handler/connection_handler_per_thread.cc:312
#15 0x0000000000eddd84 in pfs_spawn_thread (arg=0x7fc223371e20) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/storage/perfschema/pfs.cc:2190
#16 0x00007fc266805e25 in start_thread () from /lib64/libpthread.so.0
#17 0x00007fc2649d6bad in clone () from /lib64/libc.so.6
```

WORKER執行緒並不相應：
```
Thread 6 (Thread 0x7fc266d38700 (LWP 27759)):
#0  page_offset (ptr=0x7fc22b18c155) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/storage/innobase/include/page0page.ic:62
#1  page_rec_is_supremum (rec=0x7fc22b18c155 "") at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/storage/innobase/include/page0page.ic:514
#2  page_cur_is_after_last (cur=0x7fc20acb02a8) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/storage/innobase/include/page0cur.ic:142
#3  btr_pcur_is_after_last_on_page (cursor=0x7fc20acb02a0) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/storage/innobase/include/btr0pcur.ic:170
#4  btr_pcur_move_to_next (cursor=0x7fc20acb02a0, mtr=0x7fc266d37580) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/storage/innobase/include/btr0pcur.ic:353
#5  0x000000000106a819 in row_search_mvcc (buf=buf@entry=0x7fc20e3e8d30 "\377\221\\", mode=PAGE_CUR_G, mode@entry=PAGE_CUR_UNSUPP, prebuilt=0x7fc20acb0088, match_mode=match_mode@entry=0, direction=direction@entry=1) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/storage/innobase/row/row0sel.cc:6266
#6  0x0000000000f570b8 in ha_innobase::general_fetch (this=0x7fc20b069a30, buf=0x7fc20e3e8d30 "\377\221\\", direction=1, match_mode=0) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/storage/innobase/handler/ha_innodb.cc:9854
#7  0x0000000000800fcf in handler::ha_rnd_next (this=0x7fc20b069a30, buf=0x7fc20e3e8d30 "\377\221\\") at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/sql/handler.cc:3146
#8  0x0000000000e403c9 in Rows_log_event::do_table_scan_and_update (this=this@entry=0x7fc20872c2a0, rli=rli@entry=0x7fc20b466000) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/sql/log_event.cc:11096
#9  0x0000000000e3fb2e in Rows_log_event::do_apply_event (this=0x7fc20872c2a0, rli=0x7fc20b466000) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/sql/log_event.cc:11564
#10 0x0000000000e9c04f in slave_worker_exec_job_group (worker=worker@entry=0x7fc20b466000, rli=rli@entry=0x7fc2223d3000) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/sql/rpl_rli_pdb.cc:2671
#11 0x0000000000e7f42b in handle_slave_worker (arg=arg@entry=0x7fc20b466000) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/sql/rpl_slave.cc:6239
#12 0x0000000000eddd84 in pfs_spawn_thread (arg=0x7fc20b4e1220) at /mnt/workspace/percona-server-5.7-binaries-release-rocks/label_exp/min-centos-6-x64/percona-server-5.7.22-22/storage/perfschema/pfs.cc:2190
#13 0x00007fc266805e25 in start_thread () from /lib64/libpthread.so.0
#14 0x00007fc2649d6bad in clone () from /lib64/libc.so.6
```

###二、啟動流程呼叫

```
init_slave 
 ->Rpl_info_factory::create_slave_info_objects
   ->scan_repositories Rpl_info_factory::scan_repositories 掃描個數 slave_master_info 通道個數
   ->scan_repositories Rpl_info_factory::scan_repositories 掃描個數 slave_relay_log_info 通道個數
   這個函式返回是否是table型別INFO_REPOSITORY_TABLE返回通道個數
   ->load_channel_names_from_repository 調入函式使用者讀取通道名字資訊加入到channel list連結串列
     ->如果是檔案模式那麼必須是單通道並且是預設通道，加入channel list即可 加入的是channel的名字
     ->如果是表模式那麼直接透過表回去資訊，加入channel list
       ->Rpl_info_factory::load_channel_names_from_table 讀取表資訊
         內部進行迴圈讀取每行記錄加入channel的名字
   ->進入迴圈對每個channel進行操作
     ->Rpl_info_factory::create_mi_and_rli_objects 建立mi 和 rli 記憶體結構對應slave_master_info和slave_relay_log_info的通道
       ->Rpl_info_factory::create_mi
       ->Rpl_info_factory::create_rli 此處訪問了slave_worker_info表的例項個數
       ->pchannel_map->add_mi 加入到channel_map
     ->load_mi_and_rli_from_repositories 從表中讀取資料
       ->Master_info::mi_init_info 讀取slave_master_info表資訊
         ->Master_info::read_info 讀取資訊
       ->Relay_log_info::rli_init_info 讀取slave_relay_log_info表資訊 邏輯複雜
         ->初始化Retrieved_Gtid_Set 值 relay_log.init_gtid_sets relay_log_recover設定不初始化
           DBUG_PRINT("info", ("Iterating forwards through relay logs, "
                                   "updating the Retrieved_Gtid_Set and updating "
                                   "IO thread trx parser before start."));
               for (it= find(filename_list.begin(), filename_list.end(), *rit);
                    it != filename_list.end(); it++)//正向掃描relay log
               {
                 const char *filename= it->c_str();
                 DBUG_PRINT("info", ("filename='%s'", filename));
                 if (read_gtids_and_update_trx_parser_from_relaylog(filename, all_gtids,
                                                                    true, trx_parser,
                                                                    gtid_partial_trx))
                 {
                   error= 1;
                   goto end;
                 }
               }
         
         ->mysql_bin_log::set_previous_gtid_set_relaylog
         ->Relay_log_info::read_info讀取資訊
         ->如果recovery 開啟 需要呼叫init_recovery根據引數recovery 判斷是否做恢復 並且進行MTS恢復 使用rli 覆蓋 mi 同時設定rli的讀取relay位置為初始位置
           MTS recovery GTID 即便是需要 恢復也會繼續 覆蓋，如果是位點POSTION則不能進行覆蓋
          ->mts_recovery_groups
            ->MY_BITMAP *groups= &rli->recovery_groups; //在recovery 時使用的bitmap
            ->獲取lwm的位置 為CP
                 LOG_POS_COORD cp=
                 {
                       (char *) rli->get_group_master_log_name(),
                       rli->get_group_master_log_pos()
                 }; //獲取  檢查點 位置 的LWM 為CP
            ->迴圈每一個worker
              ->獲取 worker info的資訊
              ->獲取到 worker執行緒最後執行的位置 w_last
                    LOG_POS_COORD w_last= { const_cast<char>(worker->get_group_master_log_name()),
                    worker->get_group_master_log_pos() }; //獲取到 worker執行緒最後執行的位置
              ->比較cp和w_last 
                ->如果大於 則需要恢復 
                  ->將這個woker 計入臨時陣列
                ->如果小於等於則不需要恢復
            ->到這裡需要恢復的worker執行緒已經提取
            ->如果有需要恢復的worker執行緒需要初始化一個點陣圖group 
              bitmap_init(groups, NULL, MTS_MAX_BITS_IN_GROUP, FALSE);
              rli->recovery_groups_inited= true;//如果需要恢復
              bitmap_clear_all(groups);
            ->迴圈每一個需要恢復的woker
              ->獲取woker的w_last
              ->輸出日誌
                         sql_print_information("Slave: MTS group recovery relay log info based on "
                          "Worker-Id %lu, "
                          "group_relay_log_name %s, group_relay_log_pos %llu "
                          "group_master_log_name %s, group_master_log_pos %llu",
                          w->id,
                          w->get_group_relay_log_name(),
                          w->get_group_relay_log_pos(),
                          w->get_group_master_log_name(),
                          w->get_group_master_log_pos());
                          
              ->獲取checkpoint的 reply log name和relay log pos
                    if (rli->relay_log.find_log_pos(&linfo, rli->get_group_relay_log_name(), 1))

offset= rli->get_group_relay_log_pos();//讀取 RELAY POS位置
                     my_b_seek(&log, offset); 偏移到這個位置
              ->從位置開始迴圈開始掃描每一個event
                 ->如果是XID則代表是事務找到了
                   ->獲取當前ev的位置ev_coord
                   ->用ev_coord和w_last進行比較
                     ->recovery_group_cnt++說明找到了一個需要恢復的事務
                     ->輸出日誌
                                 sql_print_information("Slave: MTS group recovery relay log info "
                                "group_master_log_name %s, "
                                "event_master_log_pos %llu.",
                                rli->get_group_master_log_name(),
                                ev->common_header->log_pos);//輸出日誌
                                
                     ->如果恢復掃描完成
                        mts_event_coord_cmp(&ev_coord, &w_last)) == 0
										   ->將這個woker的 點陣圖計入到group3點陣圖
										   for (uint i= (w->checkpoint_seqno + 1) - recovery_group_cnt,
                        j= 0; i <= w->checkpoint_seqno; i++, j++)
                        {
                        if (bitmap_is_set(&w->group_executed, i))//如果這一位 已經設定
                         {
                       DBUG_PRINT("mts", ("Setting bit %u.", j));
                       bitmap_fast_test_and_set(groups, j);//那麼GTOUPS 這個 bitmap中應該設定 最終GTOUPS會包含全的需要恢復的事物
                         }
                        }
										 ->設定not_reached_commit= false;//找到了結束位置
										 ->如果沒有找到結束位置
										   sql_print_error("Error looking for file after %s.", linfo.log_file_name);
                       goto err;
										   
					   -> 最終比較每一個worker的recovery_group_cnt  及從CP開始的掃描到的事物數量
					      獲取最大值為最終值 rli->mts_recovery_group_cnt
					      rli->mts_recovery_group_cnt= (rli->mts_recovery_group_cnt < recovery_group_cnt ?                                                            
					            recovery_group_cnt : rli->mts_recovery_group_cnt);//比較每個worker的recovery_group_cnt 將最大的一個 賦值給 rli->mts_recovery_group_cnt 
                  //就是需要恢復的最大事務量     
          ->如果有需要恢復的事務
           ->是GTID且有需要恢復的事務繼續覆蓋，後面從GTID位置開始讀取恢復即可
             rli->recovery_parallel_workers= 0;//不進行恢復了
             rli->clear_mts_recovery_groups();//如果是GTID 清空 後面直接更具GTID進行恢復即可
             
          ->如果是位點則不能覆蓋直接return。後面需要填充gap                                                                                                                  
          ->recover_relay_log初始化rli info binlog位置 將使用rep 庫覆蓋 master 庫檔案 同時 將初始化
          ->繼續
       -> 如果 是POS且有需要恢復的事務 MTS 呼叫
        fill_mts_gaps_and_recover mts恢復？   
        ->啟動slave 協調執行緒進行恢復
          ->恢復完成後使用恢復到的位置開啟檔案relay 
 ->進行每個通道的檢測主要檢查GTID引數是否和通道中的通道配置一致，不一致報錯初始化引數GTID關閉但是通道是AUTO_POSITION  模式
 ->check_slave_sql_config_conflict進行衝突檢測
  -> 如果opt_slave_preserve_commit_order引數設定了，並且 slave_parallel_workers大於0 
    -> 如果是MTS_PARALLEL_TYPE_DB_NAME db模式 報錯
    -> 如果opt_bin_log或者opt_log_slave_updates沒有設定報錯
    言外之意就是如果SLAVE_PRESERVE_COMMIT_ORDER設定了並且開啟了MTS那麼必須是logic模式同時 binlog和 log_slave_updates 引數必須同時開啟
  -> 如果slave_parallel_workers大於0 並且 SLAVE_PARALLEL_TYPE 不是logic 模式 並且是 MGR的通道
    -> 那麼報錯
    
 ->如果沒有設定opt_skip_slave_start則進行啟動操作
  ->迴圈每個通道
    ->啟動每個通道 start_slave_threads 
      -> 如果設定了auto_postion 但是沒有開啟gtid模式那麼IO執行緒的啟動將會失敗
      -> mts_recovery_groups 進行MTS 恢復?再學習
      -> 根據是IO執行緒或者sql執行緒進行啟動執行緒 start_slave_thread 最終會呼叫        handle_slave_io/handle_slave_sql
```

###三、關於repository引數設定為TABLE的棧幀

我是在MTS下面的，修改使用者表棧幀
```
(gdb) bt
#0  row_update_for_mysql_using_upd_graph (mysql_rec=0x7ffe78014c58 "\376\062", prebuilt=0x7ffe78015100)
    at /root/mysqlall/percona-server-locks-detail-5.7.22/storage/innobase/row/row0mysql.cc:2955
#1  0x0000000001b3d6f1 in row_update_for_mysql (mysql_rec=0x7ffe78014c58 "\376\062", prebuilt=0x7ffe78015100)
    at /root/mysqlall/percona-server-locks-detail-5.7.22/storage/innobase/row/row0mysql.cc:3131
#2  0x00000000019d42ff in ha_innobase::update_row (this=0x7ffe780146d0, old_row=0x7ffe78014c58 "\376\062", new_row=0x7ffe78014c10 "\376\062")
    at /root/mysqlall/percona-server-locks-detail-5.7.22/storage/innobase/handler/ha_innodb.cc:9018
#3  0x0000000000f909cb in handler::ha_update_row (this=0x7ffe780146d0, old_data=0x7ffe78014c58 "\376\062", new_data=0x7ffe78014c10 "\376\062")
    at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/handler.cc:8509
#4  0x0000000001860831 in Update_rows_log_event::do_exec_row (this=0x7ffe7402f6b0, rli=0x7ffe74024a90)
    at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/log_event.cc:13322
#5  0x0000000001858059 in Rows_log_event::do_apply_row (this=0x7ffe7402f6b0, rli=0x7ffe74024a90)
    at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/log_event.cc:10395
#6  0x0000000001859f25 in Rows_log_event::do_table_scan_and_update (this=0x7ffe7402f6b0, rli=0x7ffe74024a90)
    at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/log_event.cc:11137
#7  0x000000000185b0b5 in Rows_log_event::do_apply_event (this=0x7ffe7402f6b0, rli=0x7ffe74024a90)
    at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/log_event.cc:11564
#8  0x000000000186acf2 in Log_event::do_apply_event_worker (this=0x7ffe7402f6b0, w=0x7ffe74024a90)
    at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/log_event.cc:765
#9  0x00000000018e7672 in Slave_worker::slave_worker_exec_event (this=0x7ffe74024a90, ev=0x7ffe7402f6b0)
    at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/rpl_rli_pdb.cc:1890
#10 0x00000000018e97ed in slave_worker_exec_job_group (worker=0x7ffe74024a90, rli=0x675e860)
    at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/rpl_rli_pdb.cc:2671
#11 0x00000000018c077a in handle_slave_worker (arg=0x7ffe74024a90) at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/rpl_slave.cc:6239
#12 0x0000000001945620 in pfs_spawn_thread (arg=0x7ffe74029080) at /root/mysqlall/percona-server-locks-detail-5.7.22/storage/perfschema/pfs.cc:2190
#13 0x00007ffff7bc6aa1 in start_thread () from /lib64/libpthread.so.0
#14 0x00007ffff6719bcd in clone () from /lib64/libc.so.6
(gdb) n
2956            trx_t*          trx             = prebuilt->trx;
(gdb) n
2957            ulint           fk_depth        = 0;
(gdb) n
2958            bool            got_s_lock      = false;
(gdb) n
2960            DBUG_ENTER("row_update_for_mysql_using_upd_graph");
(gdb) n
2962            ut_ad(trx);
(gdb) n
2963            ut_a(prebuilt->magic_n == ROW_PREBUILT_ALLOCATED);
(gdb) n
2964            ut_a(prebuilt->magic_n2 == ROW_PREBUILT_ALLOCATED);
(gdb) n
2967            if (prebuilt->table->ibd_file_missing) {
(gdb) n
2977            if(srv_force_recovery) {
(gdb) p trx->id
$1 = 350783
(gdb) p table->name->m_name
$2 = 0x7ffe780130f8 "testmts/tii"
(gdb) c
Continuing.
```

修改slave_worker_info棧幀

```
Breakpoint 3, row_update_for_mysql_using_upd_graph (mysql_rec=0x7ffe7800f680 "\001", prebuilt=0x7ffe78010a50)
    at /root/mysqlall/percona-server-locks-detail-5.7.22/storage/innobase/row/row0mysql.cc:2955
2955            dict_table_t*   table           = prebuilt->table;
(gdb) bt
#0  row_update_for_mysql_using_upd_graph (mysql_rec=0x7ffe7800f680 "\001", prebuilt=0x7ffe78010a50)
    at /root/mysqlall/percona-server-locks-detail-5.7.22/storage/innobase/row/row0mysql.cc:2955
#1  0x0000000001b3d6f1 in row_update_for_mysql (mysql_rec=0x7ffe7800f680 "\001", prebuilt=0x7ffe78010a50)
    at /root/mysqlall/percona-server-locks-detail-5.7.22/storage/innobase/row/row0mysql.cc:3131
#2  0x00000000019d42ff in ha_innobase::update_row (this=0x7ffe7800f020, old_row=0x7ffe7800f680 "\001", new_row=0x7ffe7800f560 "\001")
    at /root/mysqlall/percona-server-locks-detail-5.7.22/storage/innobase/handler/ha_innodb.cc:9018
#3  0x0000000000f90b70 in handler::ha_update_row (this=0x7ffe7800f020, old_data=0x7ffe7800f680 "\001", new_data=0x7ffe7800f560 "\001")
    at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/handler.cc:8509
#4  0x00000000018f7f76 in Rpl_info_table::do_flush_info (this=0x7ffe74028cf0, force=true)
    at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/rpl_info_table.cc:225
#5  0x00000000018d4376 in Rpl_info_handler::flush_info (this=0x7ffe74028cf0, force=true) at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/rpl_info_handler.h:94
#6  0x00000000018e3a8d in Slave_worker::flush_info (this=0x7ffe74024a90, force=true) at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/rpl_rli_pdb.cc:441
#7  0x00000000018e4a0e in Slave_worker::commit_positions (this=0x7ffe74024a90, ev=0x7ffe7402e688, ptr_g=0x7ffe7400e880, force=true)
    at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/rpl_rli_pdb.cc:719
#8  0x000000000184e2f6 in Xid_apply_log_event::do_apply_event_worker (this=0x7ffe7402e688, w=0x7ffe74024a90)
    at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/log_event.cc:7398
#9  0x00000000018e7672 in Slave_worker::slave_worker_exec_event (this=0x7ffe74024a90, ev=0x7ffe7402e688)
    at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/rpl_rli_pdb.cc:1890
#10 0x00000000018e97ed in slave_worker_exec_job_group (worker=0x7ffe74024a90, rli=0x675e860)
    at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/rpl_rli_pdb.cc:2671
#11 0x00000000018c077a in handle_slave_worker (arg=0x7ffe74024a90) at /root/mysqlall/percona-server-locks-detail-5.7.22/sql/rpl_slave.cc:6239
#12 0x0000000001945620 in pfs_spawn_thread (arg=0x7ffe74029080) at /root/mysqlall/percona-server-locks-detail-5.7.22/storage/perfschema/pfs.cc:2190
#13 0x00007ffff7bc6aa1 in start_thread () from /lib64/libpthread.so.0
#14 0x00007ffff6719bcd in clone () from /lib64/libc.so.6
(gdb) n
2956            trx_t*          trx             = prebuilt->trx;
(gdb) n
2957            ulint           fk_depth        = 0;
(gdb) n
2958            bool            got_s_lock      = false;
(gdb) n
2960            DBUG_ENTER("row_update_for_mysql_using_upd_graph");
(gdb) n
2962            ut_ad(trx);
(gdb) n
2963            ut_a(prebuilt->magic_n == ROW_PREBUILT_ALLOCATED);
(gdb) p trx->id
$3 = 350783
(gdb) p table->name->m_name
$4 = 0x7ffeac013f68 "mysql/slave_worker_info"
```</char></optimized></optimized></optimized></optimized></optimized></optimized>

二十五：從庫的關閉和恢復流程（筆記）

相關文章