[20220531]inactive session等待事件2.txt

来源:这里教程网 时间:2026-03-03 17:40:52 作者:

[20220531]inactive session等待事件2.txt --//检查生产系统,发现inactive session 和 enq: FU - contention 等待事件,从来没有遇到过,仔细探究看看: --//本文集中inactive session等待事件,在分析过程中走了许多弯路,重新整理. 1.环境: > @ ver1 PORT_STRING                    VERSION        BANNER ------------------------------ -------------- -------------------------------------------------------------------------------- x86_64/Linux 2.4.xx            11.2.0.4.0     Oracle Database 11g Enterprise Edition Release 11.2.0.4.0 - 64bit Production 2.ashtop查看: > @ashtop event 1=1 &day     Total                                                                                                      Distinct Distinct   Seconds     AAS %This   EVENT                          FIRST_SEEN          LAST_SEEN           Execs Seen  Tstamps --------- ------- ------- ------------------------------ ------------------- ------------------- ---------- --------    334312     3.9   93% |                                2022-05-16 10:52:20 2022-05-17 10:52:12      15444    32265     13951      .2    4% | inactive session               2022-05-16 10:52:18 2022-05-17 10:52:11       6248     8976      2854      .0    1% | enq: FU - contention           2022-05-16 11:11:39 2022-05-17 10:14:18          1     2854      2342      .0    1% | RMAN backup & recovery I/O     2022-05-16 19:30:08 2022-05-17 01:52:46          1      496      1236      .0    0% | control file sequential read   2022-05-16 10:57:37 2022-05-17 10:51:08        528      996       734      .0    0% | enq: TX - row lock contention  2022-05-16 15:43:39 2022-05-16 17:16:01          3      734 --//inactive session 和 enq: FU - contention等待事件,先集中探究inactive session,关于enq: FU - contention另外写一篇文章. > @ ev_name "inactive session" EVENT#   EVENT_ID NAME             PARAMETER1  PARAMETER2  PARAMETER3           WAIT_CLASS_ID WAIT_CLASS# WAIT_CLASS ------ ---------- ---------------- ----------- ----------- -------------------- ------------- ----------- ----------    422 3757863830 inactive session session#    waited      instance|serial         1893977003           0 Other 3.分析: > @ ashtop sql_id,machine "event='inactive session'" &day     Total                                                                                              Distinct Distinct   Seconds     AAS %This   SQL_ID        MACHINE              FIRST_SEEN          LAST_SEEN           Execs Seen  Tstamps --------- ------- ------- ------------- -------------------- ------------------- ------------------- ---------- --------         6      .0    0% | 00pxd4aar1nug fyhis2               2022-05-18 20:27:36 2022-05-18 20:27:38          2        3         6      .0    0% | 017smqud0wmq1 fyhis2               2022-05-19 09:33:28 2022-05-19 09:33:30          2        3         6      .0    0% | 01rt1d6530tsb fyhis2               2022-05-19 09:25:12 2022-05-19 09:25:14          2        3         6      .0    0% | 01tzhda22b2gw fyhis2               2022-05-19 08:19:13 2022-05-19 08:19:15          2        3         6      .0    0% | 029k77b1npypb fyhis2               2022-05-18 12:16:55 2022-05-18 12:16:57          2        3         6      .0    0% | 02qy363816jac fyhis2               2022-05-18 19:30:58 2022-05-18 19:31:00          2        3 .... 30 rows selected. > @ sql_id 00pxd4aar1nug --SQL_ID = 00pxd4aar1nug DECLARE job BINARY_INTEGER := :job; next_date DATE := :mydate;  broken BOOLEAN := FALSE; BEGIN tlogon.killsession(8481,3615); :mydate := next_date; IF broken THEN :b := 1; ELSE :b := 0; END IF; END; ; --//查看几个都是类似的语句。 --//猜测防水墙执行的kill会话的语句。格式化如下: DECLARE     job BINARY_INTEGER := :job;     next_date DATE := :mydate;     broken BOOLEAN := FALSE; BEGIN     tlogon.killsession(8481,3615);     :mydate := next_date;     IF broken THEN         :b := 1;     ELSE         :b := 0;     END IF; END; / --//猜测应该是防水墙对于无法登陆的回话kill. $ tail -f alert_ywdb1.log ORA-06512: at line 5 opiodr aborting process unknown ospid (25783) as a result of ORA-28 Tue May 17 09:11:27 2022 Errors in file /u01/app/oracle/diag/rdbms/ywdb/ywdb1/trace/ywdb1_ora_15721.trc: ORA-00604: error occurred at recursive SQL level 1 ORA-00028: your session has been killed ORA-06512: at "SYS.DBMS_LOCK", line 205 ORA-06512: at "HZMCASSET.TLOGON", line 1552 ORA-06512: at "HZMCASSET.TLOGON", line 1566 ORA-06512: at "HZMCASSET.TLOGON", line 1620 ORA-06512: at "HZMCASSET.TLOGON", line 2523 --//1566-1552 = 14 --//1620-1552 = 68 --//2523-1552 = 971 ORA-06512: at line 1 ORA-06512: at line 5 opiodr aborting process unknown ospid (15721) as a result of ORA-28 Tue May 17 09:11:30 2022 $ oerr ora 28 00028, 00000, "your session has been killed" // *Cause:  A privileged user has killed your session and you are no longer //          logged on to the database. // *Action: Login again if you wish to continue working. --//tlogon包体是加密的,不过网上很容易找到破解软件.unwrap包后,执行代码如下,注我加了行号:   94  PROCEDURE KILLSESSION(V_SID NUMBER, V_SERIAL# NUMBER) IS   95     L_SQL VARCHAR2(4000);   96   BEGIN   97     L_SQL := 'alter system kill session ' || '''' || V_SID || ',' ||   98              V_SERIAL# || '''';   99       100     EXECUTE IMMEDIATE L_SQL;  101   EXCEPTION  102     WHEN OTHERS THEN  103       ROLLBACK;  104   END; --//很简单,根据参数kill session,注意并没有immediate参数。也就是对于无法满足登陆条件的回话,直接kill. --//继续看了源代码,还有一段: 1542       PROCEDURE SUBMITKILLSESSION(V_SID NUMBER, V_SERIAL# NUMBER) IS 1543         L_JOB NUMBER; 1544         L_SQL VARCHAR2(4000); 1545       BEGIN 1546         SYS.DBMS_JOB.SUBMIT(L_JOB, 1547                             'tlogon.killsession(' || TO_CHAR(V_SID) || ',' || 1548                             TO_CHAR(V_SERIAL#) || ');', 1549                             SYSDATE, 1550                             INSTANCE => SYS_CONTEXT('userenv', 'instance')); 1551         COMMIT; 1552         SYS.DBMS_LOCK.SLEEP(3); ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1553       END; 1554     BEGIN 1555       IF TLOGONAUDIT.ACTIONLEVEL = REJECT THEN 1556         TVAR.G_1031 := TRUE; 1557         IF LOWER(SYS_CONTEXT('userenv', 'isdba')) = 'true' THEN 1558           SUBMITKILLSESSION(TLOGONAUDIT.SID, TLOGONAUDIT.SERIAL); 1559         END IF; 1560        1561         IF LOWER(SYS_CONTEXT('idctx', 'isdba')) = 'true' THEN 1562           SUBMITKILLSESSION(TLOGONAUDIT.SID, TLOGONAUDIT.SERIAL); 1563         END IF; 1564        1565         IF ISTRIGGERADMIN = TRUE THEN 1566           SUBMITKILLSESSION(TLOGONAUDIT.SID, TLOGONAUDIT.SERIAL); ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1567         END IF; 1568        1569          1570          1571         RAISE PRIVS_ERROR; 1572       END IF; 1573       TVAR.G_SESSID := TLOGONAUDIT.ID; 1574        1575     EXCEPTION 1576       WHEN OTHERS THEN 1577         L_ERRCODE := SQLCODE; 1578         L_ERRMSG  := SUBSTR(SQLERRM, 1, 100); 1579         ROLLBACK; 1580         RAISEAPPERROR(L_ERRCODE, 1581                       ERR_LOGONRULE, 1582                       STRLOGONRULE || ' doaction ' || L_ERRMSG); 1583     END; 1584    1585     FUNCTION CHECKACCOUNT RETURN BOOLEAN IS 1586     BEGIN 1587       IF SYS_CONTEXT('userenv', 'session_user') = 1588          SYS_CONTEXT('userenv', 'current_user') THEN 1589         IF TLOGONAUDIT.APPNAME LIKE 'CAPAA-%' OR 1590            TLOGONAUDIT.APPNAME = 'JOB' THEN 1591           RETURN TRUE; 1592         ELSE 1593           RETURN FALSE; 1594         END IF; 1595       END IF; 1596       RETURN TRUE; 1597     END; 1598   ... 1606     PROCEDURE DORESPONSE IS ... 1620       DOACTION; ... 2523     DORESPONSE; --//alert内容: Tue May 17 09:11:27 2022 Errors in file /u01/app/oracle/diag/rdbms/ywdb/ywdb1/trace/ywdb1_ora_15721.trc: ORA-00604: error occurred at recursive SQL level 1 ORA-00028: your session has been killed ORA-06512: at "SYS.DBMS_LOCK", line 205 ORA-06512: at "HZMCASSET.TLOGON", line 1552 ORA-06512: at "HZMCASSET.TLOGON", line 1566 ORA-06512: at "HZMCASSET.TLOGON", line 1620 ORA-06512: at "HZMCASSET.TLOGON", line 2523 --//根据alert的报错,可以推断先执行DORESPONSE,DOACTION,SUBMITKILLSESSION,然后调用SYS.DBMS_LOCK.SLEEP(3). --//我开始以为没有给SYS.DBMS_LOCK.SLEEP授权执行权限,导致出现在alert出现以上错误.仔细检查发现授权存在. --// GRANT EXECUTE ON SYS.DBMS_LOCK TO HZMCASSET; > Select job, what c40 ,  instance from DBA_JOBS where schema_user = 'HZMCASSET';        JOB C40                                        INSTANCE ---------- ---------------------------------------- ----------      61191 tlogon.killsession(581,17111);                    1      61192 tlogon.killsession(581,17111);                    1 > Select job, what c40 ,  instance from DBA_JOBS where schema_user = 'HZMCASSET'; no rows selected --//后台不断的kill这些会话。我发现有时候出现2行,有时候出现1行,为什么?我开始不理解开发为什么使用job的方式, --//直接调用KILLSESSION 过程不就可以了吗?后来才明白oracle不能kill自己.会出现如下提示: --// ORA-00027: cannot kill current session --//后来想明白了,出现2行的情况下,第一次调用job失败,然后启动另外1个job执行成功.这样前面job继续执行时正好在 --//执行SYS.DBMS_LOCK.SLEEP(3)操作(因为sleep 3秒),第2个job已经kill了回话,这样就会在alert记录如下: Tue May 17 09:11:27 2022 Errors in file /u01/app/oracle/diag/rdbms/ywdb/ywdb1/trace/ywdb1_ora_15721.trc: ORA-00604: error occurred at recursive SQL level 1 ORA-00028: your session has been killed ORA-06512: at "SYS.DBMS_LOCK", line 205 ORA-06512: at "HZMCASSET.TLOGON", line 1552 ORA-06512: at "HZMCASSET.TLOGON", line 1566 ORA-06512: at "HZMCASSET.TLOGON", line 1620 ORA-06512: at "HZMCASSET.TLOGON", line 2523 --//第一次为什么会失败,按照前面ashtop脚本的输出,可以定位出现了inactive session等待事件,阻塞了kill操作. --//为什么这样的操作会出现inactive session等待事件呢?继续分析: --//查看跟踪文件,应该是MODULE NAME:(JDBC Thin Client)时无法登录。 *** 2022-05-17 10:17:18.201 *** SESSION ID:(8757.6731) 2022-05-17 10:17:18.201 *** CLIENT ID:() 2022-05-17 10:17:18.201 *** SERVICE NAME:(ywdb) 2022-05-17 10:17:18.201 *** MODULE NAME:(JDBC Thin Client) 2022-05-17 10:17:18.201 *** ACTION NAME:() 2022-05-17 10:17:18.201 --//没有其他相关信息,也无法知道是那个用户登录 --//很奇怪应用出现错误,无法登陆,没有任何开发反馈,无语. > @ wcy &1min "event='inactive session'" -- Display ASH Wait Chain Signatures script v0.6 BETA by Tanel Poder ( http://blog.tanelpoder.com ) %This     SECONDS        AAS #Blkrs WAIT_CHAIN                                                                       FIRST_SEEN          LAST_SEEN ------ ---------- ---------- ------ -------------------------------------------------------------------------------- ------------------- -------------------   10%           3         .1      1 -> 590,9377,@2=>7628,11383,@2=>inactive session -> [idle blocker 2,590,9377]     2022-05-19 11:43:07 2022-05-19 11:43:09                                        ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~   10%           3         .1      1 -> 590,9377,@2=>295,56079,@2=>inactive session -> [idle blocker 2,590,9377]      2022-05-19 11:43:07 2022-05-19 11:43:09                                        ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~   10%           3         .1      1 -> 590,9423,@2=>295,56083,@2=>inactive session -> [idle blocker 2,590,9423]      2022-05-19 11:43:32 2022-05-19 11:43:34   10%           3         .1      1 -> 590,9423,@2=>7628,11385,@2=>inactive session -> [idle blocker 2,590,9423]     2022-05-19 11:43:32 2022-05-19 11:43:34    7%           2          0      1 -> 590,9349,@2=>3128,17545,@2=>inactive session -> [idle blocker 2,590,9349]     2022-05-19 11:42:47 2022-05-19 11:42:48    7%           2          0      1 -> 7338,30489,@1=>7340,14317,@1=>inactive session -> [idle blocker 1,7338,30489] 2022-05-19 11:43:25 2022-05-19 11:43:26    7%           2          0      1 -> 7338,30487,@1=>14,22465,@1=>inactive session -> [idle blocker 1,7338,30487]   2022-05-19 11:43:20 2022-05-19 11:43:21    7%           2          0      1 -> 7338,30489,@1=>14,22467,@1=>inactive session -> [idle blocker 1,7338,30489]   2022-05-19 11:43:25 2022-05-19 11:43:26    7%           2          0      1 -> 590,9349,@2=>7628,11379,@2=>inactive session -> [idle blocker 2,590,9349]     2022-05-19 11:42:47 2022-05-19 11:42:48    3%           1          0      1 -> ,,@=>7340,14317,@1=>inactive session                                          2022-05-19 11:43:27 2022-05-19 11:43:27    3%           1          0      1 -> 590,9351,@2=>7628,11381,@2=>inactive session -> [idle blocker 2,590,9351]     2022-05-19 11:42:52 2022-05-19 11:42:52    3%           1          0      1 -> ,,@=>7628,11379,@2=>inactive session                                          2022-05-19 11:42:49 2022-05-19 11:42:49    3%           1          0      1 -> ,,@=>14,22467,@1=>inactive session                                            2022-05-19 11:43:27 2022-05-19 11:43:27    3%           1          0      1 -> ,,@=>295,56085,@2=>inactive session                                           2022-05-19 11:43:37 2022-05-19 11:43:37    3%           1          0      1 -> ,,@=>3128,17545,@2=>inactive session                                          2022-05-19 11:42:49 2022-05-19 11:42:49    3%           1          0      1 -> ,,@=>295,56081,@2=>inactive session                                           2022-05-19 11:43:12 2022-05-19 11:43:12 16 rows selected. --//仔细看阻塞的链条.似乎阻塞链接出现在相同实例,注意看一个细节sid=7628,serial# 会变化(+2). --//噢,视乎想明白了,因为不断有用户登录,出现serial# 会变化加2的情况,另外写blog验证这个情况。 4.总结: --//通过以上分析可以得到一个结论,大量job在kill session时,会出现inactive session. --//应该可以在测试环境模拟出inactive session等待事件. --//开发应该immediate kill 会话也许就不会出现这个提示了。

相关推荐