MySQL如何找出未提交事務(wù)的SQL淺析

python_lover 2022-02-20

展開全文

很久之前曾經(jīng)總結(jié)過一篇博客“MySQL如何找出未提交事務(wù)信息”，現(xiàn)在看來，這篇文章中不少知識點或觀點都略顯膚淺，或者說不夠深入，甚至部分結(jié)論是錯誤的。下面重新探討一下這個話題。那么我們還是以之前的例子來介紹。

--準備測試環(huán)境數(shù)據(jù)（實驗環(huán)境為MySQL 8.0.18社區(qū)版）

mysql> create table kkk(id int , name varchar(12));Query OK, 0 rows affected (0.34 sec) mysql> insert into kkk values(1, 'kerry');Query OK, 1 row affected (0.01 sec) mysql> insert into kkk values(2, 'jerry');Query OK, 1 row affected (0.00 sec) mysql> insert into kkk values(3, 'ken');Query OK, 1 row affected (0.00 sec) mysql>  mysql> create table t(a   varchar(10));Query OK, 0 rows affected (0.47 sec) mysql> insert into t values('test');Query OK, 1 row affected (0.00 sec)

在一個會話窗口（連接ID=38）執(zhí)行下面SQL

mysql> select connection_id() from dual;+-----------------+| connection_id() |+-----------------+|              38 |+-----------------+1 row in set (0.00 sec) mysql> set session autocommit=0;Query OK, 0 rows affected (0.00 sec) mysql>  delete from kkk where id =1;Query OK, 1 row affected (0.00 sec) mysql> 

在另外一個會話窗口（連接ID=39）執(zhí)行下面SQL

mysql> SELECT t.trx_mysql_thread_id    ->   ,t.trx_id    ->       ,t.trx_state    ->       ,t.trx_tables_in_use    ->       ,t.trx_tables_locked    ->       ,t.trx_query    ->       ,t.trx_rows_locked     ->       ,t.trx_rows_modified    ->       ,t.trx_lock_structs    ->       ,t.trx_started    ->       ,t.trx_isolation_level    ->       ,p.time     ->       ,p.user    ->       ,p.host    ->       ,p.db    ->       ,p.command    -> FROM   information_schema.innodb_trx t     ->        INNER JOIN information_schema.processlist p     ->                ON t.trx_mysql_thread_id = p.id     -> WHERE  t.trx_state = 'RUNNING'     ->        AND p.time > 4     ->        AND p.command = 'Sleep'\G *************************** 1. row ***************************trx_mysql_thread_id: 38             trx_id: 7981581          trx_state: RUNNING  trx_tables_in_use: 0  trx_tables_locked: 1          trx_query: NULL    trx_rows_locked: 4  trx_rows_modified: 1   trx_lock_structs: 2        trx_started: 2020-12-03 15:39:37trx_isolation_level: REPEATABLE READ               time: 23               user: root               host: localhost                 db: MyDB            command: Sleep1 row in set (0.00 sec)

雖然上面這個SQL找不出事務(wù)執(zhí)行過的SQL，其實呢，MySQL中未提交事務(wù)的最后執(zhí)行的一個SQL是可以通過下面腳本準確查找出來的。如下所示：

SELECT t.trx_mysql_thread_id                        AS connection_id      ,t.trx_id                                     AS trx_id                ,t.trx_state                                  AS trx_state              ,t.trx_started                                AS trx_started           ,TIMESTAMPDIFF(SECOND,t.trx_started, now())   AS "trx_run_time(s)"        ,t.trx_requested_lock_id                      AS trx_requested_lock_id      ,t.trx_operation_state                        AS trx_operation_state      ,t.trx_tables_in_use                          AS trx_tables_in_use      ,t.trx_tables_locked                          AS trx_tables_locked      ,t.trx_rows_locked                            AS trx_rows_locked      ,t.trx_isolation_level                        AS trx_isolation_level      ,t.trx_is_read_only                           AS trx_is_read_only      ,t.trx_autocommit_non_locking                 AS trx_autocommit_non_locking      ,e.event_name                                 AS event_name      ,e.timer_wait / 1000000000000                 AS timer_wait      ,e.sql_text FROM   information_schema.innodb_trx t,        performance_schema.events_statements_current e,        performance_schema.threads c WHERE  t.trx_mysql_thread_id = c.processlist_id    AND e.thread_id = c.thread_id\G;

如下截圖所示：

在會話窗口（連接ID=38）繼續(xù)執(zhí)行下面SQL:"select * from t;"。如下所示

mysql> set session autocommit=0;Query OK, 0 rows affected (0.01 sec) mysql> delete from kkk where id =1;Query OK, 1 row affected (0.00 sec) mysql> select * from t;+------+| a    |+------+| test |+------+1 row in set (0.00 sec) mysql> 

在會話窗口（連接ID=39）上繼續(xù)執(zhí)行下面SQL，你會發(fā)現(xiàn)捕獲的是事務(wù)最后執(zhí)行的SQL語句“select * from t”

mysql> SELECT t.trx_mysql_thread_id                        AS connection_id    ->       ,t.trx_id                                     AS trx_id              ->       ,t.trx_state                                  AS trx_state            ->       ,t.trx_started                                AS trx_started         ->       ,TIMESTAMPDIFF(SECOND,t.trx_started, now())   AS "trx_run_time(s)"      ->       ,t.trx_requested_lock_id                      AS trx_requested_lock_id    ->       ,t.trx_operation_state                        AS trx_operation_state    ->       ,t.trx_tables_in_use                          AS trx_tables_in_use    ->       ,t.trx_tables_locked                          AS trx_tables_locked    ->       ,t.trx_rows_locked                            AS trx_rows_locked    ->       ,t.trx_isolation_level                        AS trx_isolation_level    ->       ,t.trx_is_read_only                           AS trx_is_read_only    ->       ,t.trx_autocommit_non_locking                 AS trx_autocommit_non_locking    ->       ,e.event_name                                 AS event_name    ->       ,e.timer_wait / 1000000000000                 AS timer_wait    ->       ,e.sql_text     -> FROM   information_schema.innodb_trx t,     ->        performance_schema.events_statements_current e,     ->        performance_schema.threads c     -> WHERE  t.trx_mysql_thread_id = c.processlist_id     ->    AND e.thread_id = c.thread_id\G; *************************** 1. row ***************************             connection_id: 38                    trx_id: 7981581                 trx_state: RUNNING               trx_started: 2020-12-03 15:39:37           trx_run_time(s): 237     trx_requested_lock_id: NULL       trx_operation_state: NULL         trx_tables_in_use: 0         trx_tables_locked: 1           trx_rows_locked: 4       trx_isolation_level: REPEATABLE READ          trx_is_read_only: 0trx_autocommit_non_locking: 0                event_name: statement/sql/select                timer_wait: 0.0002                  sql_text: select * from t1 row in set (0.00 sec) ERROR: No query specified

也是說上面SQL只能獲取未提交事務(wù)最后執(zhí)行的一個SQL語句，生產(chǎn)環(huán)境中，一個事務(wù)中往往不止一個SQL語句，而是多個SQL語句的集合。如果想將一個未提交事務(wù)里面所有執(zhí)行過的SQL找出來怎么辦呢？其實在MySQL中還是有辦法的。下面SQL語句就可以找出

SELECT ps.id                ,trx.trx_id      ,trx_started      ,trx_state      ,trx_isolation_level      ,esh.event_id      ,esh.timer_wait      ,esh.event_name       ,esh.sql_text         ,esh.returned_sqlstate      ,esh.mysql_errno      ,esh.message_text      ,esh.errors      ,esh.warnings FROM   information_schema.innodb_trx trx        JOIN information_schema.processlist ps          ON trx.trx_mysql_thread_id = ps.id        LEFT JOIN performance_schema.threads th               ON th.processlist_id = trx.trx_mysql_thread_id        LEFT JOIN performance_schema.events_statements_history esh               ON esh.thread_id = th.thread_id WHERE  trx.trx_started < current_time - INTERVAL 2 second    AND ps.user != 'SYSTEM_USER'   ORDER  BY esh.event_id;

或者

SELECT trx.trx_mysql_thread_id AS processlist_id      ,sc.thread_id      ,trx.trx_started      ,TO_SECONDS(now())-TO_SECONDS(trx_started) AS trx_last_time       ,pc1.user      ,pc1.host      ,pc1.db      ,sc.SQL_TEXT AS current_sql_text      ,sh.history_sql_testFROM INFORMATION_SCHEMA.INNODB_TRX trxINNER JOIN INFORMATION_SCHEMA.processlist pc1 ON trx.trx_mysql_thread_id=pc1.idINNER JOIN performance_schema.threads th on th.processlist_id = trx.trx_mysql_thread_idINNER JOIN performance_schema.events_statements_current sc ON sc.THREAD_ID = th.THREAD_IDINNER JOIN (                SELECT thread_id AS thread_id, GROUP_CONCAT(SQL_TEXT SEPARATOR ';') AS history_sql_test                FROM performance_schema.events_statements_history                    GROUP BY thread_id                  ) sh ON sh.thread_id = th.thread_idWHERE trx_mysql_thread_id != connection_id()  AND TO_SECONDS(now())-TO_SECONDS(trx_started) >= 0 ;

但是這兩個SQL有個問題：它會找出當前連接歷史上所有執(zhí)行過的SQL（當然前提是這些SQL還保存在performance_schema.events_statements_history表中），也就是說這個SQL，不僅查詢出未提交事務(wù)所有執(zhí)行過的腳本，而且會查詢出很多歷史SQL腳本，例如，這個會話（連接）之前事務(wù)的SQL語句，而且還有一個比較頭疼的問題：這里不好區(qū)分哪些SQL對應(yīng)哪些事務(wù)。需要借助其他信息來甄別。比較費時費力。如下截圖所示。

因為只有information_schema.innodb_trx系統(tǒng)表中包含事務(wù)的開始時間（trx_started），其它系統(tǒng)表沒有跟事務(wù)相關(guān)的時間，只能借助performance_schema.events_statements_history中的TIMER_START字段來獲取事件的SQL開始執(zhí)行的時間，而這個時間必然是大于或等于對應(yīng)事務(wù)的開始時間（trx_started）的。所以從這個突破口來找出未提交事務(wù)的所有SQL，下面是關(guān)于TIMER_START等字段的詳細介紹。

關(guān)于TIMER_START，TIMER_END，TIMER_WAIT的介紹如下：

TIMER_START，TIMER_END，TIMER_WAIT：事件的時間信息。這些值的單位是皮秒（萬億分之一秒）。

TIMER_START和TIMER_END值表示事件的開始時間和結(jié)束時間。

TIMER_WAIT是事件執(zhí)行消耗的時間（持續(xù)時間）

如果事件未執(zhí)行完成，則TIMER_END為當前時間，TIMER_WAIT為當前為止所經(jīng)過的時間（TIMER_END - TIMER_START）。

如果監(jiān)視儀器配置表setup_instruments中對應(yīng)的監(jiān)視器TIMED字段被設(shè)置為 NO，則不會收集該監(jiān)視器的時間信息，那么對于該事件采集的信息記錄中，TIMER_START，TIMER_END和TIMER_WAIT字段值均為NULL

測試、折騰了好久，終于搞出了一個幾乎完美的SQL：

SELECT @dt_ts:=UNIX_TIMESTAMP(NOW());SELECT      @dt_timer:=MAX(sh.TIMER_START)FROM performance_schema.threads AS tINNER JOIN performance_schema.events_statements_history AS shON t.`THREAD_ID`=sh.`THREAD_ID`WHERE t.PROCESSLIST_ID=CONNECTION_ID();        SELECT sh.current_schema         AS database_name      ,t.thread_id      ,it.trx_mysql_thread_id         AS connection_id      ,it.trx_id      ,sh.event_id      ,it.trx_state      ,REPLACE(REPLACE(REPLACE(sh.`SQL_TEXT`,'\n',' '),'\r',' '),'\t',' ') AS executed_sql      ,it.trx_started      ,FROM_UNIXTIME(@dt_ts-CAST((@dt_timer-sh.TIMER_START)/1000000000000  AS SIGNED)) AS start_time      ,FROM_UNIXTIME(@dt_ts-CAST((@dt_timer-sh.TIMER_END)  /1000000000000  AS SIGNED)) AS end_time      ,(sh.TIMER_END-sh.TIMER_START)/1000000000000 AS used_seconds      ,sh.TIMER_WAIT/1000000000000 AS wait_seconds      ,sh.LOCK_TIME/1000000000000 AS lock_seconds      ,sh.ROWS_AFFECTED AS affected_rows      ,sh.ROWS_SENT AS send_rowsFROM performance_schema.threads AS tINNER JOIN information_schema.innodb_trx it ON  it.trx_mysql_thread_id = t.processlist_id INNER JOIN performance_schema.events_statements_history AS sh        ON t.`THREAD_ID`=sh.`THREAD_ID`WHERE t.PROCESSLIST_ID IN (                SELECT                       p.ID AS conn_id                FROM `information_schema`.`INNODB_TRX` t                INNER JOIN `information_schema`.`PROCESSLIST` p                        ON t.trx_mysql_thread_id=p.id                WHERE t.trx_state='RUNNING'                  AND p.COMMAND='Sleep'                  AND p.TIME>2                )AND sh.TIMER_START<@dt_timerAND FROM_UNIXTIME(@dt_ts-CAST((@dt_timer-sh.TIMER_START)/1000000000000  AS SIGNED)) >=it.trx_startedORDER BY it.trx_id ASC, sh.TIMER_START ASC;

它能找出未提交事務(wù)的SQL，簡單測試完全沒有問題，同時構(gòu)造幾個未提交事務(wù)測試也OK。但是上面SQL由三個SQL組成，總感覺有點別扭，研究了一下，可以改造成下面SQL。

SELECT sh.current_schema         AS database_name      ,t.thread_id      ,it.trx_mysql_thread_id   AS connection_id      ,it.trx_id      ,sh.event_id      ,it.trx_state      ,REPLACE(REPLACE(REPLACE(sh.`SQL_TEXT`,'\n',' '),'\r',' '),'\t',' ') AS executed_sql      ,it.trx_started      ,DATE_SUB(NOW(), INTERVAL (SELECT VARIABLE_VALUE FROM performance_schema.global_status WHERE VARIABLE_NAME='UPTIME') - sh.TIMER_START*10e-13 second) AS start_time      ,DATE_SUB(NOW(), INTERVAL (SELECT VARIABLE_VALUE FROM performance_schema.global_status WHERE VARIABLE_NAME='UPTIME') - sh.TIMER_END*10e-13 second)   AS end_time      ,(sh.TIMER_END-sh.TIMER_START)/1000000000000 AS used_seconds      ,sh.TIMER_WAIT/1000000000000 AS wait_seconds      ,sh.LOCK_TIME/1000000000000 AS lock_seconds      ,sh.ROWS_AFFECTED AS affected_rows      ,sh.ROWS_SENT AS send_rowsFROM performance_schema.threads AS tINNER JOIN information_schema.innodb_trx it ON  it.trx_mysql_thread_id = t.processlist_id INNER JOIN performance_schema.events_statements_history AS sh        ON t.`THREAD_ID`=sh.`THREAD_ID`WHERE t.PROCESSLIST_ID IN (                SELECT                       p.ID AS conn_id                FROM `information_schema`.`INNODB_TRX` t                INNER JOIN `information_schema`.`PROCESSLIST` p                        ON t.trx_mysql_thread_id=p.id                WHERE t.trx_state='RUNNING'                  AND p.COMMAND='Sleep'                  AND p.TIME>2                )AND sh.TIMER_START<(SELECT VARIABLE_VALUE*1000000000000 FROM performance_schema.global_status WHERE VARIABLE_NAME='UPTIME')AND DATE_SUB(NOW(), INTERVAL (SELECT VARIABLE_VALUE FROM performance_schema.global_status WHERE VARIABLE_NAME='UPTIME') - sh.TIMER_START*10e-13 second) >=it.trx_startedORDER BY it.trx_id ASC, sh.TIMER_START ASC;

注意：performance_schema.global_status是MySQL 5.7引入的，如果數(shù)據(jù)庫是MySQL 5.6的話，可以用INFORMATION_SCHEMA.GLOBAL_STATUS替換SQL中的performance_schema.global_status

那么是否這個SQL就一定完美了呢？網(wǎng)友MSSQL123反饋在一個測試環(huán)境中，發(fā)現(xiàn)上面這種SQL居然查不出任何數(shù)據(jù)，因為FROM_UNIXTIME(@dt_ts-CAST((@dt_timer-sh.TIMER_START)/1000000000000 AS SIGNED)) >=it.trx_started會將數(shù)據(jù)過濾掉，檢查發(fā)現(xiàn)對應(yīng)的trx_started值都大于start_time

-------------------------------------------------------------------------補充資料--------------------------------------------------------------------------------------------------------

那么同一個測試環(huán)境，隔天測試的時候，突然發(fā)現(xiàn)上面第一個SQL正常，第二個SQL就由于不同的寫法，start_time有細微的差別，導致查詢結(jié)果完全不同（第二個SQL語句精確到毫秒，對比的時候由于偏差過濾掉一批數(shù)據(jù)）

------------------------------------------------------------------------------------------------------------------------------------------------

搜索相關(guān)文檔，了解到TIMER_START字段值可能有波動或偏差，那么這個波動或偏差可能影響查詢結(jié)果，下面內(nèi)容來自http://porthos.ist./docs/mySQL/performance-schema.html

Modifications to the setup_timers table affect monitoring immediately. Events already in progress may use the original timer for the begin time and the new timer for the end time. To avoid unpredictable results after you make timer changes, use TRUNCATE TABLE to reset Performance Schema statistics.

The timer baseline (“time zero”) occurs at Performance Schema initialization during server startup. TIMER_START and TIMER_END values in events represent picoseconds since the baseline. TIMER_WAIT values are durations in picoseconds.

Picosecond values in events are approximate. Their accuracy is subject to the usual forms of error associated with conversion from one unit to another. If the CYCLE timer is used and the processor rate varies, there might be drift. For these reasons, it is not reasonable to look at the TIMER_START value for an event as an accurate measure of time elapsed since server startup. On the other hand, it is reasonable to use TIMER_START or TIMER_WAIT values in ORDER BY clauses to order events by start time or duration.

The choice of picoseconds in events rather than a value such as microseconds has a performance basis. One implementation goal was to show results in a uniform time unit, regardless of the timer. In an ideal world this time unit would look like a wall-clock unit and be reasonably precise; in other words, microseconds. But to convert cycles or nanoseconds to microseconds, it would be necessary to perform a division for every instrumentation. Division is expensive on many platforms. Multiplication is not expensive, so that is what is used. Therefore, the time unit is an integer multiple of the highest possible TIMER_FREQUENCY value, using a multiplier large enough to ensure that there is no major precision loss. The result is that the time unit is “picoseconds.” This precision is spurious, but the decision enables overhead to be minimized.

Before MySQL 5.7.8, while a wait, stage, statement, or transaction event is executing, the respective current-event tables display the event with TIMER_START populated, but with TIMER_END and TIMER_WAIT set to NULL

其中一段內(nèi)容翻譯如下：

事件中的皮秒值是近似值。它們的準確性受與從一個單位轉(zhuǎn)換到另一單位相關(guān)的常見誤差形式的影響。如果使用了CYCLE定時器，并且處理器速率有所變化，則可能會有偏差。由于這些原因，將事件的TIMER_START值視為自服務(wù)器啟動以來經(jīng)過的時間的準確度量是不合理的。另一方面，在ORDER BY子句中使用TIMER_START或TIMER_WAIT值來按開始時間或持續(xù)時間對事件進行排序是合理的。

我們往往想一擊必殺的解決問題，但是由于復雜的環(huán)境和一些不可控因素，現(xiàn)實往往就是“沒有銀彈”這么殘酷。如果遇到TIMER_START的波動或偏差影響查詢結(jié)果時，那么我們必須通過文章前面的SQL找出大量SQL，然后通過其它字段或信息人工甄別哪些才是未提交事務(wù)的SQL。