mysql历史数据自动归档

数据库跑一段时间后，因为查询性能、磁盘容量，运维管理等方面的原因，需要将在线数据挪到历史库(不同的服务器）。如我们的在线订单只留3个月数据，3个月以前的就需要到历史库查了。

自动归档常见的方式有pt-archiver，但我还是觉得自己写存储过程更靠谱。。。

思路：

• 在线库实例打开federated支持，创建数据库dborder(业务库), linkhis(归档用);
• 历史库创建历史表dborderhis.myorder_tab_his;
• 在linkhis库下创建federated表linkhis.myorder_tab_his，指向dborderhis.myorder_tab_his；
• 在linkhis库下创建日志表archive_log，存储过程proc_archive，proc_archive_pkg，并通过JOB调度proc_archive_pkg；
• 在线库的从库需要忽略linkhis的复制：replicate-ignore-db=linkhis，否则从库也会往这个历史库重复同步数据。

日志记录表archive_log

create table archive_log
(
  id bigint auto_increment PRIMARY key,
  tab_name           varchar(40),
  archive_date_begin datetime,
  archive_date_end   datetime,
  create_time        datetime default CURRENT_TIMESTAMP(),
  status             int(1),
  insert_rows        bigint(11),
  delete_rows        bigint(11),
  remark             varchar(1000)
)

存储过程：proc_archive

CREATE PROCEDURE proc_archive(in i_table_source varchar(40), 
                              in i_table_target varchar(40), 
                              in i_fieldname varchar(40), 
                              in i_keepdays int,
                              in i_archdays int,
                              in i_other_cond varchar(500))
begin
  /*
  入参：
  i_table_source:原表,含dbname
  i_table_target:federated表
  i_fieldname:时间字段
  i_keepdays:保留天数
  i_archdays:每次归档多少天数据
  i_other_cond:数据额外条件(如status in (2,3)不能归档，需要保留),无额外条件则输入'1=1'
  归档日志表archive_log.status字段含义：
  0:成功, 1:现有数据在保留天数内, 2:目标表含有待归档时间范围的数据, 
  3:插入数据和删除数据记录数不同, 4:SQL执行异常，具体错误见remark
  注意：
  有额外条件时，如果历史数据被修改，从不符合归档条件变成符合归档条件，
  因历史表中归档时间段内已经有之前归档的数据(@v_his_num_before>0)，程序会退出，需手动处理
  */
  declare EXIT HANDLER for SQLWARNING,NOT FOUND,SQLEXCEPTION 
  begin 
      GET DIAGNOSTICS CONDITION 1 @p1=RETURNED_SQLSTATE,@p2= MESSAGE_TEXT;
      ROLLBACK;
      insert into archive_log(tab_name,archive_date_begin,archive_date_end,status,insert_rows,delete_rows,remark)
      values(i_table_source,@v_arch_begin,@v_arch_end,4,@v_his_num_after,@v_del_num,concat('error ',@p1,' - ',@p2));
  end;
  /* 获取在线表的最小日期 */
  set @mystmt = concat("select str_to_date(date_format(min(",i_fieldname,"),'%Y%m%d'),'%Y%m%d') into @v_arch_begin from ",i_table_source,' where ',i_other_cond);
  prepare stmt from @mystmt;
  execute stmt;
  deallocate prepare stmt;

  set @v_arch_end = date_add(@v_arch_begin,interval i_archdays day);

  set @mystmt = concat("select count(*) into @v_his_num_before from ",i_table_target," where ",i_fieldname," >= ? and ",i_fieldname," < ?");
  prepare stmt from @mystmt;
  execute stmt using @v_arch_begin,@v_arch_end;
  deallocate prepare stmt;
  /* 如果在线表的数据低于keepday范围，退出 */
  if timestampdiff(day,@v_arch_begin,now()) <= i_keepdays then
      insert into archive_log(tab_name,archive_date_begin,archive_date_end,status,insert_rows,delete_rows,remark)
      values(i_table_source,@v_arch_begin,@v_arch_end,1,0,0,concat('error, all data in keey days, min ',i_fieldname,': ',@v_arch_begin));
  end if;
  /* 如果历史表所在的日期区间有数据，退出（需要手动排查原因） */
  if @v_his_num_before <> 0 then
      insert into archive_log(tab_name,archive_date_begin,archive_date_end,status,insert_rows,delete_rows,remark)
      values(i_table_source,@v_arch_begin,@v_arch_end,2,0,0,concat('error, data exists,row num:',@v_his_num_before));
  end if;
  
  if (timestampdiff(day,@v_arch_begin,now()) > i_keepdays and @v_his_num_before = 0) then 
      set @mystmt = concat("insert into ",i_table_target," select * from ",i_table_source," where ",i_fieldname," >= ? and ",i_fieldname," < ? and ",i_other_cond);
      prepare stmt from @mystmt;
      execute stmt using @v_arch_begin,@v_arch_end;
      deallocate prepare stmt;
      /* 因为federated引擎不支持事务，数据insert后再select下记录数，与下面的delete记录数对比，相同则提交delete操作 */
      set @mystmt = concat("select count(*) into @v_his_num_after from ",i_table_target," where ",i_fieldname," >= ? and ",i_fieldname," < ?");
      prepare stmt from @mystmt;
      execute stmt using @v_arch_begin,@v_arch_end;
      deallocate prepare stmt;

      start transaction;

      set @mystmt = concat("delete from ",i_table_source," where ",i_fieldname," >= ? and ",i_fieldname," < ? and ",i_other_cond);
      prepare stmt from @mystmt;
      execute stmt using @v_arch_begin,@v_arch_end;
      set @v_del_num = row_count();
      deallocate prepare stmt;

      if @v_del_num = @v_his_num_after then
          commit;
          insert into archive_log(tab_name,archive_date_begin,archive_date_end,status,insert_rows,delete_rows,remark)
          values(i_table_source,@v_arch_begin,@v_arch_end,0,@v_his_num_after,@v_del_num,'success');
      else
          rollback;
          insert into archive_log(tab_name,archive_date_begin,archive_date_end,status,insert_rows,delete_rows,remark)
          values(i_table_source,@v_arch_begin,@v_arch_end,3,@v_his_num_after,@v_del_num,'rollback, inserted rows num not equal to deleted rows num');
      end if;
    end if;
        
end;

 

存储过程proc_archive_pkg

CREATE PROCEDURE `proc_archive_pkg`()
begin
call proc_archive(
                  'dborder.myorder_tab', -- tabel source
                  'myorder_tab_his',     -- table target
                  'create_time',         -- time field name
                  120,                   -- i_keepdays
                  1,                     -- i_archdays
                  '1=1'                  -- i_other_cond
                  );

end;

归档日志表记录

最后一步就是用zabbix监控这个表的状态了。