x$kccle视图深入剖析
今天是2014-05-27,实在无聊顺便研究一下x$kccle的内容吧。如下全部是自己分析和实验结果,真实可靠。 1、如何获得v$log的底层表?我们可以通过autotrace完成查看如: SQL set autotrace trace expSQL set linesize 200SQL select * from v$log;Execution Pl
今天是2014-05-27,实在无聊顺便研究一下x$kccle的内容吧。如下全部是自己分析和实验结果,真实可靠。
1、如何获得v$log的底层表?我们可以通过autotrace完成查看如:
SQL> set autotrace trace exp SQL> set linesize 200 SQL> select * from v$log; Execution Plan ---------------------------------------------------------- Plan hash value: 2536105608 -------------------------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | -------------------------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | 1 | 197 | 0 (0)| 00:00:01 | | 1 | NESTED LOOPS | | 1 | 197 | 0 (0)| 00:00:01 | |* 2 | FIXED TABLE FULL | X$KCCLE | 1 | 148 | 0 (0)| 00:00:01 | |* 3 | FIXED TABLE FIXED INDEX| X$KCCRT (ind:1) | 1 | 49 | 0 (0)| 00:00:01 | -------------------------------------------------------------------------------------------- Predicate Information (identified by operation id): --------------------------------------------------- 2 - filter("LE"."LEDUP"0 AND "LE"."INST_ID"=USERENV('INSTANCE')) 3 - filter("RT"."INST_ID"=USERENV('INSTANCE') AND "LE"."LETHR"="RT"."RTNUM") SQL>
可以知道当我们在查看v$log视图的时候,其实就是查看的x$kccle视图字段信息。查看该视图如下:
SQL> desc x$kccle Name Null? Type ----------------------------------------- -------- ---------------------------- ADDR RAW(8) INDX NUMBER INST_ID NUMBER 实例号 LENUM NUMBER 类似group_number LESIZ NUMBER logfile的size,是操作系统的blocks数目 LESEQ NUMBER v$log中的SEQUENCE# LEHWS NUMBER LEBSZ NUMBER block size(操作系统的块) LENAB NUMBER LEFLG NUMBER 表示日志的状态信息 LETHR NUMBER v$log中的THREAD# LELFF NUMBER 下一个logfile LELFB NUMBER 前一个logfile LELOS VARCHAR2(16) low scn类似v$log FIRST_CHANGE# LELOT VARCHAR2(20) low time类似v$log FIRST_TIME LENXS VARCHAR2(16) nex scn 类似v$log中下一个logfile的low FIRST_CHANGE#相同 LENXT VARCHAR2(20) nex time类似v$log中下一个logfile的low FIRST_TIME 相同 LEPVS VARCHAR2(16) prev scn, 与上一个logfile的low scn LEARF NUMBER archive link forward LEARB NUMBER archive link backward LEFNH NUMBER LEFNT NUMBER LEDUP NUMBER 等同于v$log中members SQL>
下面对比一下:
SQL> select * from v$log; GROUP# THREAD# SEQUENCE# BYTES MEMBERS ARC STATUS FIRST_CHANGE# FIRST_TIME ---------- ---------- ---------- ---------- ---------- --- ---------------- ------------- ------------ 1 1 2944 104857600 1 YES INACTIVE 121424676 21-MAY-14 2 1 2945 104857600 1 NO CURRENT 121427428 21-MAY-14 3 1 2943 104857600 1 YES INACTIVE 121421889 21-MAY-14 SQL> SQL> select inst_id,lenum,lesiz,leseq,lehws,lebsz,lebsz*lesiz log_size,lenab,leflg,lethr,lelff,lelfb,lelos from x$kccle; INST_ID LENUM LESIZ LESEQ LEHWS LEBSZ LOG_SIZE LENAB LEFLG LETHR LELFF LELFB LELOS ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------------- 1 1 204800 2944 2 512 104857600 190256 1 1 2 0 121424676 1 2 204800 2945 1 512 104857600 4294967295 8 1 3 1 121427428 1 3 204800 2943 2 512 104857600 190256 1 1 0 2 121421889 SQL> SQL> select inst_id,lenum,leseq,lethr,lelot,lenxs,lenxt,lepvs,learf,learb,lefnh,lefnt,ledup from x$kccle; INST_ID LENUM LESEQ LETHR LELOT LENXS LENXT LEPVS LEARF LEARB LEFNH LEFNT LEDUP ---------- ----- ----- ---------- -------------------- ---------------- -------------------- ---------------- ---------- ---------- ---------- ---------- ---------- 1 1 2944 1 05/21/2014 11:26:42 121427428 05/21/2014 11:26:55 121421889 0 0 1 1 1 1 2 2945 1 05/21/2014 11:26:55 281474976710655 121424676 0 0 2 2 1 1 3 2943 1 05/21/2014 11:26:30 121424676 05/21/2014 11:26:42 121419052 0 0 3 3 1 SQL>
现在单独关注一下leflg字段,当该值为8说明是正在使用的logfile group:
dump出控制文件中的log file信息如下:
SQL> select * from v$log; GROUP# THREAD# SEQUENCE# BYTES BLOCKSIZE MEMBERS ARC STATUS FIRST_CHANGE# FIRST_TIM NEXT_CHANGE# NEXT_TIME ---------- ---------- ---------- ---------- ---------- ---------- --- ---------------- ------------- --------- ------------ --------- 1 1 19 52428800 512 2 NO CURRENT 1013368 27-MAY-14 2.8147E+14 2 1 17 52428800 512 2 NO ACTIVE 1013348 27-MAY-14 1013365 27-MAY-14 3 1 18 52428800 512 2 NO ACTIVE 1013365 27-MAY-14 1013368 27-MAY-14 SQL> oradebug setmypid Statement processed. SQL> oradebug dump controlf 12 Statement processed.
注意:也可以使用alter system set events ‘immediate trace name controlf level 12';完成控制文件的dump操作:
LOG FILE RECORDS *************************************************************************** (size = 72, compat size = 72, section max = 16, section in-use = 3, last-recid= 17, old-recno = 0, last-recno = 0) (extent = 1, blkno = 10, numrecs = 16) LOG FILE #1: name #5: /u01/app/oracle/oradata/RHYS/onlinelog/o1_mf_1_9r1z70w0_.log name #6: /u01/app/oracle/fast_recovery_area/RHYS/onlinelog/o1_mf_1_9r1z70z2_.log Thread 1 redo log links: forward: 2 backward: 0 siz: 0x19000 seq: 0x00000013 hws: 0x1 bsz: 512 nab: 0xffffffff flg: 0x8 dup: 2 Archive links: fwrd: 0 back: 0 Prev scn: 0x0000.000f7675 Low scn: 0x0000.000f7678 05/27/2014 17:54:08 Next scn: 0xffff.ffffffff 01/01/1988 00:00:00 LOG FILE #2: name #3: /u01/app/oracle/oradata/RHYS/onlinelog/o1_mf_2_9r1z726s_.log name #4: /u01/app/oracle/fast_recovery_area/RHYS/onlinelog/o1_mf_2_9r1z7291_.log Thread 1 redo log links: forward: 3 backward: 1 siz: 0x19000 seq: 0x00000011 hws: 0x2 bsz: 512 nab: 0x14 flg: 0x0 dup: 2 Archive links: fwrd: 0 back: 0 Prev scn: 0x0000.000f75ec Low scn: 0x0000.000f7664 05/27/2014 17:54:01 Next scn: 0x0000.000f7675 05/27/2014 17:54:05 LOG FILE #3: name #1: /u01/app/oracle/oradata/RHYS/onlinelog/o1_mf_3_9r1z73jf_.log name #2: /u01/app/oracle/fast_recovery_area/RHYS/onlinelog/o1_mf_3_9r1z73lp_.log Thread 1 redo log links: forward: 0 backward: 2 siz: 0x19000 seq: 0x00000012 hws: 0x2 bsz: 512 nab: 0x4 flg: 0x0 dup: 2 Archive links: fwrd: 0 back: 0 Prev scn: 0x0000.000f7664 Low scn: 0x0000.000f7675 05/27/2014 17:54:05 Next scn: 0x0000.000f7678 05/27/2014 17:54:08
对应一下:
SQL> select inst_id,lenum,leseq,lethr,lelot,lenxs,lenxt,lepvs,learf,learb,lefnh,lefnt,ledup from x$kccle; INST_ID LENUM LESEQ LETHR LELOT LENXS LENXT LEPVS LEARF LEARB LEFNH LEFNT LEDUP ---------- ----- ----- ---------- -------------------- ---------------- -------------------- ---------------- ---------- ---------- ---------- ---------- ---------- 1 1 19 1 05/27/2014 17:54:08 281474976710655 1013365 0 0 5 6 2 1 2 17 1 05/27/2014 17:54:01 1013365 05/27/2014 17:54:05 1013228 0 0 3 4 2 1 3 18 1 05/27/2014 17:54:05 1013368 05/27/2014 17:54:08 1013348 0 0 1 2 2 SQL> select inst_id,lenum,lesiz,leseq,lehws,lebsz,lebsz*lesiz log_size,lenab,leflg,lethr,lelff,lelfb,lelos from x$kccle; INST_ID LENUM LESIZ LESEQ LEHWS LEBSZ LOG_SIZE LENAB LEFLG LETHR LELFF LELFB LELOS ---------- ----- ---------- ----- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------------- 1 1 102400 19 1 512 52428800 4294967295 8 1 2 0 1013368 1 2 102400 17 2 512 52428800 20 0 1 3 1 1013348 1 3 102400 18 2 512 52428800 4 0 1 0 2 1013365 SQL>
拿group3做一个分析:如下:
LOG FILE #3:(代表第三个group)
name #1: /u01/app/oracle/oradata/RHYS/onlinelog/o1_mf_3_9r1z73jf_.log(第一个日志文件)
name #2: /u01/app/oracle/fast_recovery_area/RHYS/onlinelog/o1_mf_3_9r1z73lp_.log(第二个日志文件)
Thread 1(线程1x$kccle.lethr) redo log links: forward: 0(对应x$kccle的LELFF为0) backward: 2(对应x$kccle的lelfb为2)
siz: 0x19000(对应x$kccle的lesiz转为10进制正好是102400): seq: 0x00000012(转为10进制为18对应x$kccle的leseq): hws: 0x2(对应x$kccle的lehws为2) bsz: 512(数据块对应x$kccle的lebsz)
nab: 0x4(对应lenab为4) flg: 0x0(对应x$kccle的leflg) dup: 2(对应x$kccle的ledup)
Archive links: fwrd: 0(前一个链表x$kccle的learf) back: 0 (对应x$kccle的learb)Prev scn: 0x0000.000f7664(转为10进制scn为1013348,对应lepvs)
Low scn: 0x0000.000f7675 05/27/2014 17:54:05(对应x$kccle的lelos)
Next scn: 0x0000.000f7678 05/27/2014 17:54:08(对应x$kccle的lenxt)
由此可知x$kccle正式抓取的是控制文件中的内容,这也是为什么数据块在mount状态下,可以查看v$log的原因了。
上一篇: 浅析PHP创建ZIP档案文件技巧
下一篇: PHP开发规范