【摘要】對於mysql主備例項，seconds_behind_master是衡量master與slave之間延時的一個重要引數。通過在slave上執行"show slave status;"可以獲取seconds_behind_master的值。
Seconds_Behind_Master
對於mysql主備例項，seconds_behind_master是衡量master與slave之間延時的一個重要引數。通過在slave上執行"show slave status;"可以獲取seconds_behind_master的值。
原始實現
Definition：The number of seconds that the slave SQL thread is behind processing the master binary log.
Type：time_t（long）
計算方式如下：
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)); }
主要分為以下兩種情況：
• SQL執行緒等待IO執行緒獲取主機binlog，此時seconds_behind_master為0，表示備機與主機之間無延時；
• SQL執行緒處理relay log，此時seconds_behind_master通過(long)(time(0) – mi->rli->last_master_timestamp) – mi->clock_diff_with_master計算得到；
last_master_timestamp
定義：
• 主庫binlog中事件的時間。
• type: time_t (long)
計算方式：
last_master_timestamp根據備機是否並行複製有不同的計算方式。
非並行複製：
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); }
在該模式下，last_master_timestamp表示為每一個event的結束時間，其中when.tv_sec表示event的開始時間，exec_time表示事務的執行時間。該值的計算在apply_event之前，所以event還未執行時，last_master_timestamp已經被更新。由於exec_time僅在Query_log_event中存在，所以last_master_timestamp在應用一個事務的不同event階段變化。以一個包含兩條insert語句的事務為例，在該程式碼段的呼叫時，列印出event的型別、時間戳和執行時間
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
並行複製：
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" */
在該模式下備機上存在一個分發佇列gaq，如果gaq為空，則設定last_commit_timestamp為0；如果gaq不為空，則此時維護一個checkpoint點lwm，lwm之前的事務全部在備機上執行完成，此時last_commit_timestamp被更新為lwm所在事務執行完成後的時間。該時間型別為time_t型別。
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); }
在並行複製下，event執行完成之後才會更新last_master_timestamp，所以非並行複製和並行複製下的seconds_behind_master會存在差異。
clock_diff_with_master
定義：
• 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); }
該差值僅被計算一次，在master與slave建立聯絡時處理。
其他
exec_time
定義：
• 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();
時間函式
（1）time_t time(time_t timer) time_t為long型別，返回的數值僅精確到秒；
（2）int gettimeofday (struct timeval *tv, struct timezone *tz) 可以獲得微秒級的當前時間；
（3）timeval結構
#include <time.h> stuct timeval { time_t tv_sec; /*seconds*/ suseconds_t tv_usec; /*microseconds*/ }
總結
使用seconds_behind_master衡量主備延時只能精確到秒級別，且在某些場景下，seconds_behind_master並不能準確反映主備之間的延時。主備異常時，可以結合seconds_behind_master原始碼進行具體分析。
