欢迎您访问程序员文章站本站旨在为大家提供分享程序员计算机编程知识!
您现在的位置是: 首页  >  IT编程

详解MySQL的Seconds_Behind_Master

程序员文章站 2022-03-07 12:29:42
seconds_behind_master对于mysql主备实例,seconds_behind_master是衡量master与slave之间延时的一个重要参数。通过在slave上执行"show sl...

seconds_behind_master

对于mysql主备实例,seconds_behind_master是衡量master与slave之间延时的一个重要参数。通过在slave上执行"show slave status;"可以获取seconds_behind_master的值。

原始实现

definition:the number of seconds that the slave sql thread is behind processing the master binary log.

type:time_t(long)

计算方式如下:

rpl_slave.cc::show_slave_status_send_data()
if ((mi->get_master_log_pos() == mi->rli->get_group_master_log_pos()) &&
       (!strcmp(mi->get_master_log_name(),
                mi->rli->get_group_master_log_name()))) {
     if (mi->slave_running == mysql_slave_run_connect)
       protocol->store(0ll);
     else
       protocol->store_null();
   } else {
     long time_diff = ((long)(time(0) - mi->rli->last_master_timestamp) -
                       mi->clock_diff_with_master);
     protocol->store(
         (longlong)(mi->rli->last_master_timestamp ? max(0l, time_diff) : 0));
   }

主要分为以下两种情况:

  • sql线程等待io线程获取主机binlog,此时seconds_behind_master为0,表示备机与主机之间无延时;
  • sql线程处理relay log,此时seconds_behind_master通过(long)(time(0) – mi->rli->last_master_timestamp) – mi->clock_diff_with_master计算得到;

last_master_timestamp

定义:

主库binlog中事件的时间。

type: time_t (long)

计算方式:

last_master_timestamp根据备机是否并行复制有不同的计算方式。

非并行复制:

rpl_slave.cc:exec_relay_log_event()
if ((!rli->is_parallel_exec() || rli->last_master_timestamp == 0) &&
    !(ev->is_artificial_event() || ev->is_relay_log_event() ||
     (ev->common_header->when.tv_sec == 0) ||
     ev->get_type_code() == binary_log::format_description_event ||
     ev->server_id == 0))
{
 rli->last_master_timestamp= ev->common_header->when.tv_sec +
                             (time_t) ev->exec_time;
 dbug_assert(rli->last_master_timestamp >= 0);
}

在该模式下,last_master_timestamp表示为每一个event的结束时间,其中when.tv_sec表示event的开始时间,exec_time表示事务的执行时间。该值的计算在apply_event之前,所以event还未执行时,last_master_timestamp已经被更新。由于exec_time仅在query_log_event中存在,所以last_master_timestamp在应用一个事务的不同event阶段变化。以一个包含两条insert语句的事务为例,在该代码段的调用时,打印出event的类型、时间戳和执行时间

create table t1(a int primary key auto_increment ,b longblob) engine=innodb;
begin;
insert into t1(b) select repeat('a',104857600);
insert into t1(b) select repeat('a',104857600);
commit;

10t06:41:32.628554z 11 [note] [my-000000] [repl] event_type: 33 gtid_log_event

2020-02-10t06:41:32.628601z 11 [note] [my-000000] [repl] event_time: 1581316890

2020-02-10t06:41:32.628614z 11 [note] [my-000000] [repl] event_exec_time: 0

2020-02-10t06:41:32.628692z 11 [note] [my-000000] [repl] event_type: 2   query_event

2020-02-10t06:41:32.628704z 11 [note] [my-000000] [repl] event_time: 1581316823

2020-02-10t06:41:32.628713z 11 [note] [my-000000] [repl] event_exec_time: 35

2020-02-10t06:41:32.629037z 11 [note] [my-000000] [repl] event_type: 19   table_map_event

2020-02-10t06:41:32.629057z 11 [note] [my-000000] [repl] event_time: 1581316823

2020-02-10t06:41:32.629063z 11 [note] [my-000000] [repl] event_exec_time: 0

2020-02-10t06:41:33.644111z 11 [note] [my-000000] [repl] event_type: 30    write_rows_event

2020-02-10t06:41:33.644149z 11 [note] [my-000000] [repl] event_time: 1581316823

2020-02-10t06:41:33.644156z 11 [note] [my-000000] [repl] event_exec_time: 0

2020-02-10t06:41:43.520272z 0 [note] [my-011953] [innodb] page cleaner took 9185ms to flush 3 and evict 0 pages

2020-02-10t06:42:05.982458z 11 [note] [my-000000] [repl] event_type: 19   table_map_event

2020-02-10t06:42:05.982488z 11 [note] [my-000000] [repl] event_time: 1581316858

2020-02-10t06:42:05.982495z 11 [note] [my-000000] [repl] event_exec_time: 0

2020-02-10t06:42:06.569345z 11 [note] [my-000000] [repl] event_type: 30    write_rows_event

2020-02-10t06:42:06.569376z 11 [note] [my-000000] [repl] event_time: 1581316858

2020-02-10t06:42:06.569384z 11 [note] [my-000000] [repl] event_exec_time: 0

2020-02-10t06:42:16.506176z 0 [note] [my-011953] [innodb] page cleaner took 9352ms to flush 8 and evict 0 pages

2020-02-10t06:42:37.202507z 11 [note] [my-000000] [repl] event_type: 16    xid_event

2020-02-10t06:42:37.202539z 11 [note] [my-000000] [repl] event_time: 1581316890

2020-02-10t06:42:37.202546z 11 [note] [my-000000] [repl] event_exec_time: 0

并行复制:

rpl_slave.cc   mts_checkpoint_routine
ts = rli->gaq->empty()
          ? 0
          : reinterpret_cast<slave_job_group *>(rli->gaq->head_queue())->ts;
 rli->reset_notified_checkpoint(cnt, ts, true);
 /* end-of "coordinator::"commit_positions" */

在该模式下备机上存在一个分发队列gaq,如果gaq为空,则设置last_commit_timestamp为0;如果gaq不为空,则此时维护一个checkpoint点lwm,lwm之前的事务全部在备机上执行完成,此时last_commit_timestamp被更新为lwm所在事务执行完成后的时间。该时间类型为time_t类型。

ptr_group->ts = common_header->when.tv_sec +
                   (time_t)exec_time;  // seconds_behind_master related
rli->rli_checkpoint_seqno++;
if (update_timestamp) {
 mysql_mutex_lock(&data_lock);
 last_master_timestamp = new_ts;
 mysql_mutex_unlock(&data_lock);
}

在并行复制下,event执行完成之后才会更新last_master_timestamp,所以非并行复制和并行复制下的seconds_behind_master会存在差异。

clock_diff_with_master

定义:

  • the difference in seconds between the clock of the master and the clock of the slave (second - first). it must be signed as it may be <0 or >0. clock_diff_with_master is computed when the i/o thread starts; for this the i/o thread does a select unix_timestamp() on the master.
  • type: long
rpl_slave.cc::get_master_version_and_clock()
if (!mysql_real_query(mysql, string_with_len("select unix_timestamp()")) &&
     (master_res= mysql_store_result(mysql)) &&
     (master_row= mysql_fetch_row(master_res)))
 {
   mysql_mutex_lock(&mi->data_lock);
   mi->clock_diff_with_master=
     (long) (time((time_t*) 0) - strtoul(master_row[0], 0, 10));
   dbug_execute_if("dbug.mts.force_clock_diff_eq_0",
     mi->clock_diff_with_master= 0;);
   mysql_mutex_unlock(&mi->data_lock);
 }

该差值仅被计算一次,在master与slave建立联系时处理。

其他

exec_time

定义:

  • the difference from the statement's original start timestamp and the time at which it completed executing.
  • type: unsigned long
struct timeval end_time;
ulonglong micro_end_time = my_micro_time();
my_micro_time_to_timeval(micro_end_time, &end_time);
exec_time = end_time.tv_sec - thd_arg->query_start_in_secs();

时间函数

(1)time_t time(time_t timer) time_t为long类型,返回的数值仅精确到秒;

(2)int gettimeofday (struct timeval *tv, struct timezone *tz) 可以获得微秒级的当前时间;

(3)timeval结构

#include <time.h>
stuct timeval {
   time_t tv_sec; /*seconds*/
   suseconds_t tv_usec; /*microseconds*/
}

总结

使用seconds_behind_master衡量主备延时只能精确到秒级别,且在某些场景下,seconds_behind_master并不能准确反映主备之间的延时。主备异常时,可以结合seconds_behind_master源码进行具体分析。

以上就是详解mysql的seconds_behind_master的详细内容,更多关于mysql seconds_behind_master的资料请关注其它相关文章!