MySQL DDL简析（1）：inplace DDL 主要st

作者: 重庆八怪 | 来源:发表于2021-10-21 15:34 被阅读0次


  if (new_stage == &srv_stage_alter_table_read_pk_internal_sort) {
    m_cur_phase = READ_PK;
  } else if (new_stage == &srv_stage_alter_table_merge_sort) {
    m_cur_phase = SORT;
  } else if (new_stage == &srv_stage_alter_table_insert) {
    m_cur_phase = INSERT;
  } else if (new_stage == &srv_stage_alter_table_flush) {
    m_cur_phase = FLUSH;
  } else if (new_stage == &srv_stage_alter_table_log_index) {
    m_cur_phase = LOG_INDEX;
  } else if (new_stage == &srv_stage_alter_table_log_table) {
    m_cur_phase = LOG_TABLE;
  } else if (new_stage == &srv_stage_alter_table_end) {
    m_cur_phase = END;
  } else {
    ut_error;
  }

ha_innobase::inplace_alter_table_impl
 ->row_merge_build_indexes
  ->进入状态
     srv_stage_alter_table_read_pk_internal_sort
  ->row_merge_read_clustered_index
     -> trx->op_info = "reading clustered index"
        事务状态更改为reading clustered index
     -> 获取innodb_tmpdir 参数设置
        thd_innodb_tmpdir
     -> for (ulint i = 0; i < n_index; i++)
        循环每一个索引为其建立buffer
        merge_buf[i] = row_merge_buf_create(index[i])
        ->读取为innodb_sort_buffer_size 默认4M
     -> 从头开始读取主键 
        btr_pcur_open_at_index_side
     -> 循环每一个page，获取一个记录进行处理
        page_cur_move_to_next
        获取记录的游标
        page_cur_is_after_last(cur)
        是否是page的最后一条记录，使用supremum进行判定
        ->需要判断，是否占用了index lock
          rw_lock_get_waiters(dict_index_get_lock(clust_index)
          ->是，btr_pcur_move_to_prev_on_page(&pcur)
            先回复游标到上一次page
            os_thread_yield
            先暂时放弃CPU
          ->否，btr_page_get_next
            访问下一个page
            next_page_no == FIL_NULL
            如果下一个page为空则结束
        rec = page_cur_get_rec(cur);
        读取记录
        offsets = rec_get_offsets(rec, clust_index, nullptr, ULINT_UNDEFINED, &row_heap); 
        获取行的偏移量 
        if (online)
          是online，则
          trx->read_view->changes_visible 
          则需要改变行的可变性
          rec_get_deleted_flag
          如果是标记为删除的记录，则跳过
          不是online，则
          rec_get_deleted_flag
          如果是标记为删除的记录，则跳过
        row = row_build_w_add_vcol，实际调用为row_build_low
          建立新行，根据新建的语法，如果只是增加索引这里
          add_cols=0x0, add_v=0x0, col_map=0x0
          可以看到add cols没有具体的信息
          返回信息为dtupe_t类型，为一行根据新字段建立好的行
          已经转变为tupe类型
        ->for (ulint i = 0; i < n_nonnull; i++)
          循环所有的NOT NULL字段
          dfield_is_null(field)
          判断是否字段对应的值为NULL
          是则进行抛错处理
        ->是否需要设置自增值
          设置自增值
        ->skip_sort = skip_pk_sort && merge_buf[0]->index->is_clustered(); 
          如果为增加字段（非instant）第一次 cluster 不需要进行排序，
          如果增加的是单个索引则这里为 false，因为不是主键
        ->for (ulint i = 0; i < n_index; i++, skip_sort = false)
          循环每个需要新建的索引,这里注意条件skip_sort = false 
          除了cluster索引外其他的二级索引都需要排序
          
          ->row_merge_buf_add（通常返回0代表超过了srv_sort_buf_size大小，正常为1行处理完成） 
           对新加入的索引和现有行数据进行处理，并且写入到merge buffer中
           ->循环每一个字段
             ->拷贝field
             ->获取field长度
             ->data_size += len
               计算data总的长度
           判断总的data长度不能大于srv_sort_buf_size的大小
           如果buf->total_size + data_size >= srv_sort_buf_size - 1
             是，则返回0（multi-value index特殊处理），也就是(row_merge_buf_t)buf中的大小
             加上本次的数据大小大于了srv_sort_buf_size，就会返回
           ->循环每一个字段
             将数据从field拷贝到buf->heap
           通常返回0代表超过了srv_sort_buf_size大小，正常为1行处理完成
          ->通过row_merge_buf_add的返回值进行判断
            ->如果返回为1，则进行下一轮循环，这里的循环是循环下一次需要重建的索引
            ->如果返回为0，则说明sort buffer满了需要进行排序了
          ->是否需要跳过排序（主键才会）
            ->是
              ->row_merge_spatial_rows 和空间索引相关不考虑直接return
              ->row_merge_insert_index_tuples
                直接将记录插入到索引，跳过排序（一般这里是主键）
                改变状态为srv_stage_alter_table_insert （alter table (insert)）
                显然这个状态会多次进入，如果有多个索引
                ->for (;;)
                循环sort buffer中的每行 
                退出条件为行数大于了innobase sort buffer中的行数
                ->row_merge_mtuple_to_dtuple
                  将每行转换为tuple格式
                ->btr_bulk->insert(dtuple)
                  将dtuple以bulk方式插入，这是一个重点的行数
                  代表的了自底而上的插入方式。这里先跳过
              ->清空innobase sort buffer
            ->否
              ->是否是唯一索引
                ->是
                  需要开启重复值判定
                  row_merge_buf_sort(buf, &dup);
                  进行排序
                ->否
                  row_merge_buf_sort(buf, nullptr);
                  进行排序
                -> if (row == nullptr && file->fd == -1 && !clust_temp_file) 
                  是否开启了临时文件，如果开启了说明需要写到
                  临时文件                
                  ->否
                    row_merge_insert_index_tuples
                    排序后进行插入,如上状态alter table (insert)
                  ->是
                    row_merge_file_create_if_needed
                    建立临时文件
                    row_merge_buf_write /tmp目录需要文件 写入block
                    row_merge_write(debug一下这里) /tmp目录需要文件 将block写入到merge文件
                    写入到临时文件，后面做merge排序合并   
  ->循环每个需要建立的索引
   ->merge_files[i].fd >= 0
    先判断是否有临时文件，如果没有则不需要merge_sort和insert_index_tuples
    因为前面已经插入了  
    ->row_merge_sort 输入为前面得到的tmpfd，也就是包含数据的临时文件
     大数据量表做merge排序，这里改变状态为 srv_stage_alter_table_merge_sort
    ->row_merge_insert_index_tuples
     见后面。是有bulk方式插入到索引，这里改变状态为 srv_stage_alter_table_insert
    ->row_merge_file_destroy(&merge_files[i])
     丢弃临时文件，可以看到这里online DDL通常
     需要临时文件的支持
   ->判断是否重建主键，如果是进行下面步骤
    ->FlushObserver::flush
      ->进入状态srv_stage_alter_table_flush
      ->buf_flush_get_dirty_pages_count
       ->循环每个pool，对操作的对象的space id查看
         需要刷新块
      ->buf_LRU_flush_or_remove_pages
       类型为BUF_REMOVE_FLUSH_WRITE
       ->根据类型判断是否需要维护AHI
         BUF_REMOVE_FLUSH_WRITE不需要维护AHI
       ->buf_LRU_flush_or_remove_pages
         类型为BUF_REMOVE_FLUSH_WRITE
         进行flush刷脏，需要回写脏数据到物理文件
         flush to disk if true but don't remove
         可见代码buf_flush_or_remove_page
         根据flush与否是否从buffer pool中去掉
    ->row_merge_write_redo
      写入redo MLOG_INDEX_LOAD
    ->row_log_apply
      ->stage->begin_phase_log_index()
        进入状态srv_stage_alter_table_log_index
      ->应用日志
  ->清理merge排序用到的临时文件
  先要判断是否重建了主键ctx->need_rebuild()
  如果是
  ->row_log_table_apply 
  进入状态：srv_stage_alter_table_log_table
  ->rw_lock_x_lock(dict_index_get_lock(clust_index));  
   先对主键index加锁
  ->row_log_table_apply_ops 
   分3层，循环1：
   循环1:访问一个block(sort buffer大小)
   ->判断是否当前block(sort buffer 大小)已经是最后一个block
     如果是则直接加锁应用最后一个block，同时这种情况肯定就在
     sort buffer内存了。
   ->如果不是则需要读取物理文件，读取block(sort buffer大小)同时解锁
     ，不影响正常的写入。
     ->循环2：
       获取block中的每行记录，应用。
       ->row_log_table_apply_op
         ->根据不同类型ROW_T_INSERT/ROW_T_DELETE/ROW_T_UPDATE
           分别应用，这里以insert为例
           ->row_log_table_apply_insert
             ->row_log_table_apply_convert_mrec
               转日志记录为tuple类型最好插入准备
             ->row_log_table_apply_insert_low
               ->row_ins_clust_index_entry_low
                 插入为先主键插入
               ->循环3：
                 ->row_ins_sec_index_entry_low
                 循环每个二级索引插入
  ->rw_lock_x_unlock(dict_index_get_lock(clust_index)); 
    主键解锁

8.0 增加索引没有看到 data目录下的临时文件
增加索引和add col一起看到了，并且很慢，全索引merge，
3倍空间，data临时文件 tmp临时文件。

-rw-r-----. 1 mysql mysql 335544320 Oct 17 23:43 tttoo.ibd
-rw-r-----. 1 mysql mysql 92274688 Oct 17 23:49 #sql-ib1339-3955111351.ibd

mysql> alter table tttoo add index(io);
Query OK, 0 rows affected, 1 warning (1 min 1.46 sec)
Records: 0 Duplicates: 0 Warnings: 1

mysql> alter table tttoo add index(io),add ipp int;
Query OK, 0 rows affected, 1 warning (7 min 52.48 sec)
Records: 0 Duplicates: 0 Warnings: 1

row_log_tmpfile 为打开临时文件的接口

lrwx------. 1 root root 64 Oct 20 11:43 60 -> /newdata/mysql/mysql8023/tmp/#284070 (deleted)
lrwx------. 1 root root 64 Oct 20 11:43 59 -> /newdata/mysql/mysql8023/tmp/#284069 (deleted)
lrwx------. 1 root root 64 Oct 20 11:43 58 -> /newdata/mysql/mysql8023/tmp/#284068 (deleted)
lrwx------. 1 root root 64 Oct 20 11:43 57 -> /newdata/mysql/mysql8023/tmp/#284067 (deleted)
lrwx------. 1 root root 64 Oct 20 11:43 56 -> /newdata/mysql/mysql8023/tmp/#284066 (deleted)
lrwx------. 1 root root 64 Oct 20 11:43 55 -> /newdata/mysql/mysql8023/tmp/#284065 (deleted)
lrwx------. 1 root root 64 Oct 20 11:43 54 -> /newdata/mysql/mysql8023/tmp/#284059 (deleted)
lrwx------. 1 root root 64 Oct 20 11:43 51 -> /newdata/mysql/mysql8023/data/test/#sql-ib1394-938072227.ibd
lrwx------. 1 root root 64 Oct 20 11:43 50 -> /newdata/mysql/mysql8023/data/test/tttoo.ibd
lrwx------. 1 root root 64 Oct 20 11:43 49 -> socket:[42365]
lrwx------. 1 root root 64 Oct 20 11:43 48 -> /newdata/mysql/mysql8023/data/test/test.ibd
lrwx------. 1 root root 64 Oct 20 11:43 46 -> socket:[35028]
lrwx------. 1 root root 64 Oct 20 11:46 52 -> socket:[42886]
(gdb) p log->fd
$4 = 54

[root@mgr4 fd]# lsof -p 5517|grep delete
mysqld  5517 mysql   10u      REG               8,16      4856   262456 /newdata/mysql/mysql8023/tmp/#262456 (deleted)
mysqld  5517 mysql   11u      REG               8,16         0   262460 /newdata/mysql/mysql8023/tmp/#262460 (deleted)
mysqld  5517 mysql   12u      REG               8,16         0   262478 /newdata/mysql/mysql8023/tmp/#262478 (deleted)
mysqld  5517 mysql   18u      REG               8,16         0   262479 /newdata/mysql/mysql8023/tmp/#262479 (deleted)
mysqld  5517 mysql   54u      REG               8,16         0   284059 /newdata/mysql/mysql8023/tmp/#284059 (deleted)
mysqld  5517 mysql   55u      REG               8,16  29360128   284065 /newdata/mysql/mysql8023/tmp/#284065 (deleted)
mysqld  5517 mysql   56u      REG               8,16  29360128   284066 /newdata/mysql/mysql8023/tmp/#284066 (deleted)
mysqld  5517 mysql   57u      REG               8,16  29360128   284067 /newdata/mysql/mysql8023/tmp/#284067 (deleted)
mysqld  5517 mysql   58u      REG               8,16  29360128   284068 /newdata/mysql/mysql8023/tmp/#284068 (deleted)
mysqld  5517 mysql   59u      REG               8,16  29360128   284069 /newdata/mysql/mysql8023/tmp/#284069 (deleted)
mysqld  5517 mysql   60u      REG               8,16  29360128   284070 /newdata/mysql/mysql8023/tmp/#284070 (deleted)

/newdata/mysql/mysql8023/tmp/#284059：为log日志文件，注意这些文件都是以sort参数大小为block大小的

[root@mgr4 fd]# lsof -p 5517|grep delete
mysqld  5517 mysql   10u      REG               8,16      4856   262456 /newdata/mysql/mysql8023/tmp/#262456 (deleted)
mysqld  5517 mysql   11u      REG               8,16         0   262460 /newdata/mysql/mysql8023/tmp/#262460 (deleted)
mysqld  5517 mysql   12u      REG               8,16         0   262478 /newdata/mysql/mysql8023/tmp/#262478 (deleted)
mysqld  5517 mysql   18u      REG               8,16         0   262479 /newdata/mysql/mysql8023/tmp/#262479 (deleted)
mysqld  5517 mysql   54u      REG               8,16   4194304   284059 /newdata/mysql/mysql8023/tmp/#284059 (deleted)
mysqld  5517 mysql   56u      REG               8,16  29360128   284066 /newdata/mysql/mysql8023/tmp/#284066 (deleted)
mysqld  5517 mysql   57u      REG               8,16  29360128   284067 /newdata/mysql/mysql8023/tmp/#284067 (deleted)
mysqld  5517 mysql   58u      REG               8,16  29360128   284068 /newdata/mysql/mysql8023/tmp/#284068 (deleted)
mysqld  5517 mysql   59u      REG               8,16  29360128   284069 /newdata/mysql/mysql8023/tmp/#284069 (deleted)
mysqld  5517 mysql   60u      REG               8,16  29360128   284070 /newdata/mysql/mysql8023/tmp/#284070 (deleted)



(gdb) bt
#0  row_ins_clust_index_entry_low (flags=23, mode=33, index=0xc675ea0, n_uniq=1, entry=0x1f684778, thr=0xc4e8998, dup_chk_only=false)
    at /newdata/mysql-8.0.23/storage/innobase/row/row0ins.cc:2350
#1  0x000000000515c01c in row_log_table_apply_insert_low (thr=0xc4e8998, row=0x1f684630, trx_id=447979, offsets_heap=0x1f688670, heap=0x1f684590, dup=0x7fffe47c8980)
    at /newdata/mysql-8.0.23/storage/innobase/row/row0log.cc:1503
#2  0x000000000515c327 in row_log_table_apply_insert (thr=0xc4e8998, mrec=0x7fff9c3f9003 "", offsets=0xc6a0360, offsets_heap=0x1f688670, heap=0x1f684590, dup=0x7fffe47c8980, trx_id=447979)
    at /newdata/mysql-8.0.23/storage/innobase/row/row0log.cc:1587
#3  0x000000000515e8ea in row_log_table_apply_op (thr=0xc4e8998, trx_id_col=1, new_trx_id_col=1, dup=0x7fffe47c8980, error=0x7fffe47c8594, offsets_heap=0x1f688670, heap=0x1f684590, 
    mrec=0x7fff9c3f9003 "", mrec_end=0x7fff9c7f9000 "", offsets=0xc6a0360) at /newdata/mysql-8.0.23/storage/innobase/row/row0log.cc:2390
#4  0x00000000051604e6 in row_log_table_apply_ops (thr=0xc4e8998, dup=0x7fffe47c8980, stage=0xc6a0560) at /newdata/mysql-8.0.23/storage/innobase/row/row0log.cc:2925
#5  0x00000000051609a2 in row_log_table_apply (thr=0xc4e8998, old_table=0xc341770, table=0xc4e5490, stage=0xc6a0560) at /newdata/mysql-8.0.23/storage/innobase/row/row0log.cc:3014
#6  0x0000000004fb2b5e in ha_innobase::inplace_alter_table_impl<dd::Table> (this=0xc33f6d8, altered_table=0xc4e5490, ha_alter_info=0x7fffe47c9510, old_dd_tab=0xc33dd78, 
    new_dd_tab=0xc4d1b08) at /newdata/mysql-8.0.23/storage/innobase/handler/handler0alter.cc:6143
#7  0x0000000004f8d1b3 in ha_innobase::inplace_alter_table (this=0xc33f6d8, altered_table=0xc4e5490, ha_alter_info=0x7fffe47c9510, old_dd_tab=0xc33dd78, new_dd_tab=0xc4d1b08)
    at /newdata/mysql-8.0.23/storage/innobase/handler/handler0alter.cc:1284
#8  0x00000000038e1942 in handler::ha_inplace_alter_table (this=0xc33f6d8, altered_table=0xc4e5490, ha_alter_info=0x7fffe47c9510, old_table_def=0xc33dd78, new_table_def=0xc4d1b08)
    at /newdata/mysql-8.0.23/sql/handler.h:5785
#9  0x00000000038c5b36 in mysql_inplace_alter_table (thd=0xad622d0, schema=..., new_schema=..., table_def=0xc33dd78, altered_table_def=0xc4d1b08, table_list=0xc4c9c48, table=0xc359a90, 
    altered_table=0xc4e5490, ha_alter_info=0x7fffe47c9510, inplace_supported=HA_ALTER_INPLACE_NO_LOCK_AFTER_PREPARE, alter_ctx=0x7fffe47ca3f0, columns=std::set with 0 elements, 
    fk_key_info=0xc514b80, fk_key_count=0, fk_invalidator=0x7fffe47ca320) at /newdata/mysql-8.0.23/sql/sql_table.cc:13008
#10 0x00000000038d13eb in mysql_alter_table (thd=0xad622d0, new_db=0xc4ca260 "test", new_name=0x0, create_info=0x7fffe47cc210, table_list=0xc4c9c48, alter_info=0x7fffe47cc0a0)
    at /newdata/mysql-8.0.23/sql/sql_table.cc:16910
#11 0x0000000003e778de in Sql_cmd_alter_table::execute (this=0xc4ca500, thd=0xad622d0) at /newdata/mysql-8.0.23/sql/sql_alter.cc:351
#12 0x00000000037fa060 in mysql_execute_command (thd=0xad622d0, first_level=true) at /newdata/mysql-8.0.23/sql/sql_parse.cc:4407
#13 0x00000000037fbf41 in dispatch_sql_command (thd=0xad622d0, parser_state=0x7fffe47cda50) at /newdata/mysql-8.0.23/sql/sql_parse.cc:4988
#14 0x00000000037f2543 in dispatch_command (thd=0xad622d0, com_data=0x7fffe47ceb00, command=COM_QUERY) at /newdata/mysql-8.0.23/sql/sql_parse.cc:1836
#15 0x00000000037f095e in do_command (thd=0xad622d0) at /newdata/mysql-8.0.23/sql/sql_parse.cc:1320
#16 0x00000000039c5c91 in handle_connection (arg=0xacee820) at /newdata/mysql-8.0.23/sql/conn_handler/connection_handler_per_thread.cc:301
#17 0x000000000562cd84 in pfs_spawn_thread (arg=0xab1c320) at /newdata/mysql-8.0.23/storage/perfschema/pfs.cc:2900
#18 0x00007ffff7bc6ea5 in start_thread () from /lib64/libpthread.so.0
#19 0x00007ffff5e388dd in clone () from /lib64/libc.so.6


#0  row_ins_sec_index_entry_low (flags=23, mode=33, index=0xc6771f0, offsets_heap=0x1f688670, heap=0x1f684590, entry=0x1f684960, trx_id=447979, thr=0xc4e8998, dup_chk_only=false)
    at /newdata/mysql-8.0.23/storage/innobase/row/row0ins.cc:2775
#1  0x000000000515c14a in row_log_table_apply_insert_low (thr=0xc4e8998, row=0x1f684630, trx_id=447979, offsets_heap=0x1f688670, heap=0x1f684590, dup=0x7fffe47c8980)
    at /newdata/mysql-8.0.23/storage/innobase/row/row0log.cc:1532
#2  0x000000000515c327 in row_log_table_apply_insert (thr=0xc4e8998, mrec=0x7fff9c3f9003 "", offsets=0xc6a0360, offsets_heap=0x1f688670, heap=0x1f684590, dup=0x7fffe47c8980, trx_id=447979)
    at /newdata/mysql-8.0.23/storage/innobase/row/row0log.cc:1587
#3  0x000000000515e8ea in row_log_table_apply_op (thr=0xc4e8998, trx_id_col=1, new_trx_id_col=1, dup=0x7fffe47c8980, error=0x7fffe47c8594, offsets_heap=0x1f688670, heap=0x1f684590, 
    mrec=0x7fff9c3f9003 "", mrec_end=0x7fff9c7f9000 "", offsets=0xc6a0360) at /newdata/mysql-8.0.23/storage/innobase/row/row0log.cc:2390
#4  0x00000000051604e6 in row_log_table_apply_ops (thr=0xc4e8998, dup=0x7fffe47c8980, stage=0xc6a0560) at /newdata/mysql-8.0.23/storage/innobase/row/row0log.cc:2925
#5  0x00000000051609a2 in row_log_table_apply (thr=0xc4e8998, old_table=0xc341770, table=0xc4e5490, stage=0xc6a0560) at /newdata/mysql-8.0.23/storage/innobase/row/row0log.cc:3014
#6  0x0000000004fb2b5e in ha_innobase::inplace_alter_table_impl<dd::Table> (this=0xc33f6d8, altered_table=0xc4e5490, ha_alter_info=0x7fffe47c9510, old_dd_tab=0xc33dd78, 
    new_dd_tab=0xc4d1b08) at /newdata/mysql-8.0.23/storage/innobase/handler/handler0alter.cc:6143
#7  0x0000000004f8d1b3 in ha_innobase::inplace_alter_table (this=0xc33f6d8, altered_table=0xc4e5490, ha_alter_info=0x7fffe47c9510, old_dd_tab=0xc33dd78, new_dd_tab=0xc4d1b08)
    at /newdata/mysql-8.0.23/storage/innobase/handler/handler0alter.cc:1284
#8  0x00000000038e1942 in handler::ha_inplace_alter_table (this=0xc33f6d8, altered_table=0xc4e5490, ha_alter_info=0x7fffe47c9510, old_table_def=0xc33dd78, new_table_def=0xc4d1b08)
    at /newdata/mysql-8.0.23/sql/handler.h:5785
#9  0x00000000038c5b36 in mysql_inplace_alter_table (thd=0xad622d0, schema=..., new_schema=..., table_def=0xc33dd78, altered_table_def=0xc4d1b08, table_list=0xc4c9c48, table=0xc359a90, 
    altered_table=0xc4e5490, ha_alter_info=0x7fffe47c9510, inplace_supported=HA_ALTER_INPLACE_NO_LOCK_AFTER_PREPARE, alter_ctx=0x7fffe47ca3f0, columns=std::set with 0 elements, 
    fk_key_info=0xc514b80, fk_key_count=0, fk_invalidator=0x7fffe47ca320) at /newdata/mysql-8.0.23/sql/sql_table.cc:13008
#10 0x00000000038d13eb in mysql_alter_table (thd=0xad622d0, new_db=0xc4ca260 "test", new_name=0x0, create_info=0x7fffe47cc210, table_list=0xc4c9c48, alter_info=0x7fffe47cc0a0)
    at /newdata/mysql-8.0.23/sql/sql_table.cc:16910
#11 0x0000000003e778de in Sql_cmd_alter_table::execute (this=0xc4ca500, thd=0xad622d0) at /newdata/mysql-8.0.23/sql/sql_alter.cc:351
#12 0x00000000037fa060 in mysql_execute_command (thd=0xad622d0, first_level=true) at /newdata/mysql-8.0.23/sql/sql_parse.cc:4407
#13 0x00000000037fbf41 in dispatch_sql_command (thd=0xad622d0, parser_state=0x7fffe47cda50) at /newdata/mysql-8.0.23/sql/sql_parse.cc:4988
#14 0x00000000037f2543 in dispatch_command (thd=0xad622d0, com_data=0x7fffe47ceb00, command=COM_QUERY) at /newdata/mysql-8.0.23/sql/sql_parse.cc:1836
#15 0x00000000037f095e in do_command (thd=0xad622d0) at /newdata/mysql-8.0.23/sql/sql_parse.cc:1320
#16 0x00000000039c5c91 in handle_connection (arg=0xacee820) at /newdata/mysql-8.0.23/sql/conn_handler/connection_handler_per_thread.cc:301
#17 0x000000000562cd84 in pfs_spawn_thread (arg=0xab1c320) at /newdata/mysql-8.0.23/storage/perfschema/pfs.cc:2900
#18 0x00007ffff7bc6ea5 in start_thread () from /lib64/libpthread.so.0
#19 0x00007ffff5e388dd in clone () from /lib64/libc.so.6

row lock记录的类型。
row_log_table_delete
row_log_table_update
row_log_table_insert

delete记录比较少，主键+undoptr+DATA_TRX_ID_LEN = 没有主键就是20字节


log过大ER_INNODB_ONLINE_LOG_TOO_BIG

应用最后一部分：

#0  row_log_table_apply_ops (thr=0xc4e8998, dup=0x7fffe47c80a0, stage=0xc6a0560) at /newdata/mysql-8.0.23/storage/innobase/row/row0log.cc:2661
#1  0x00000000051609a2 in row_log_table_apply (thr=0xc4e8998, old_table=0xc341770, table=0xc4e5490, stage=0xc6a0560) at /newdata/mysql-8.0.23/storage/innobase/row/row0log.cc:3014
#2  0x0000000004facbe2 in commit_try_rebuild (ha_alter_info=0x7fffe47c9510, ctx=0xc515b98, altered_table=0xc4e5490, old_table=0xc359a90, trx=0x7fffe5be9078, table_name=0xc34a3e5 "tttoo")
    at /newdata/mysql-8.0.23/storage/innobase/handler/handler0alter.cc:6836
#3  0x0000000004fb3c09 in ha_innobase::commit_inplace_alter_table_impl<dd::Table> (this=0xc33f6d8, altered_table=0xc4e5490, ha_alter_info=0x7fffe47c9510, commit=true, 
    old_dd_tab=0xc33dd78, new_dd_tab=0xc4d1b08) at /newdata/mysql-8.0.23/storage/innobase/handler/handler0alter.cc:7473
#4  0x0000000004f8d34d in ha_innobase::commit_inplace_alter_table (this=0xc33f6d8, altered_table=0xc4e5490, ha_alter_info=0x7fffe47c9510, commit=true, old_dd_tab=0xc33dd78, 
    new_dd_tab=0xc4d1b08) at /newdata/mysql-8.0.23/storage/innobase/handler/handler0alter.cc:1331
#5  0x0000000003b77f6b in handler::ha_commit_inplace_alter_table (this=0xc33f6d8, altered_table=0xc4e5490, ha_alter_info=0x7fffe47c9510, commit=true, old_table_def=0xc33dd78, 
    new_table_def=0xc4d1b08) at /newdata/mysql-8.0.23/sql/handler.cc:4879
#6  0x00000000038c5cfc in mysql_inplace_alter_table (thd=0xad622d0, schema=..., new_schema=..., table_def=0xc33dd78, altered_table_def=0xc4d1b08, table_list=0xc4c9c48, table=0xc359a90, 
    altered_table=0xc4e5490, ha_alter_info=0x7fffe47c9510, inplace_supported=HA_ALTER_INPLACE_NO_LOCK_AFTER_PREPARE, alter_ctx=0x7fffe47ca3f0, columns=std::set with 0 elements, 
    fk_key_info=0xc514b80, fk_key_count=0, fk_invalidator=0x7fffe47ca320) at /newdata/mysql-8.0.23/sql/sql_table.cc:13050
#7  0x00000000038d13eb in mysql_alter_table (thd=0xad622d0, new_db=0xc4ca260 "test", new_name=0x0, create_info=0x7fffe47cc210, table_list=0xc4c9c48, alter_info=0x7fffe47cc0a0)
    at /newdata/mysql-8.0.23/sql/sql_table.cc:16910
#8  0x0000000003e778de in Sql_cmd_alter_table::execute (this=0xc4ca500, thd=0xad622d0) at /newdata/mysql-8.0.23/sql/sql_alter.cc:351
#9  0x00000000037fa060 in mysql_execute_command (thd=0xad622d0, first_level=true) at /newdata/mysql-8.0.23/sql/sql_parse.cc:4407
#10 0x00000000037fbf41 in dispatch_sql_command (thd=0xad622d0, parser_state=0x7fffe47cda50) at /newdata/mysql-8.0.23/sql/sql_parse.cc:4988
#11 0x00000000037f2543 in dispatch_command (thd=0xad622d0, com_data=0x7fffe47ceb00, command=COM_QUERY) at /newdata/mysql-8.0.23/sql/sql_parse.cc:1836
#12 0x00000000037f095e in do_command (thd=0xad622d0) at /newdata/mysql-8.0.23/sql/sql_parse.cc:1320
#13 0x00000000039c5c91 in handle_connection (arg=0xacee820) at /newdata/mysql-8.0.23/sql/conn_handler/connection_handler_per_thread.cc:301
#14 0x000000000562cd84 in pfs_spawn_thread (arg=0xab1c320) at /newdata/mysql-8.0.23/storage/perfschema/pfs.cc:2900
#15 0x00007ffff7bc6ea5 in start_thread () from /lib64/libpthread.so.0
#16 0x00007ffff5e388dd in clone () from /lib64/libc.so.6

MySQL DDL简析（1）：inplace DDL 主要st

相关文章

网友评论

延伸阅读

深度阅读

栏目导航

热点阅读