oracle segment undo_71_UNDO扩展学习
UNDO擴展學習
UNDO是Oracle在UNDO SEGMENT(回滾段)中記錄的信息。下面使用UNDO這名字可能會包含兩種意思，一是前面說的回滾段中的記錄、二是UNDO整個機制。
在這部分內容里，沒有涉及閃回技術，要知道不少閃回技術是基于UNDO機制來實現的，比如flashback query、flashback table、flashback version query等。
首先看下我的操作環境：
sys@L10GR204> select * from v$version;
BANNER
----------------------------------------------------------------------------------------------------
Oracle Database 10g Enterprise Edition Release 10.2.0.4.0 - 64bi
PL/SQL Release 10.2.0.4.0 - Production
CORE 10.2.0.4.0 Production
TNS for Linux: Version 10.2.0.4.0 - Production
NLSRTL Version 10.2.0.4.0 - Production
下面列出本節要討論的內容：
1) 什么時候聯機回滾段？
2) 如何證明SYSTEM回滾段只對系統對象提供服務
3) alter system set undo_tablespace = uuu;
4) 如何創建一個新的UNDO表空間并使用
5) 回滾段增減
6) 了解幾個參數
7) 構造一個CR塊
8) 快速構造ora-01555 快照過舊
9) enq: US – contention
10) Oracle INSERT、UPDATE、DELETE與Undo
什么時候聯機回滾段？
這是一個非常簡單的問題，如果大家已經掌握了數據庫的啟動過程，那就很容易回答上了。
看下面簡單實驗：
當前實例為MOOUNT狀態
sys@L10GR204> select status from v$instance;
STATUS
------------------------
MOUNTED
這時候試圖訪問和UNDO相關視圖，將會報錯
sys@L10GR204> select * from v$rollname;
select * from v$rollname
*
ERROR at line 1:
ORA-01219: database not open: queries allowed on fixed tables/views only
當我們OPEN之后，即可讀取回滾段信息
sys@L10GR204> alter database open;
Database altered.
sys@L10GR204> col name for a55
sys@L10GR204> select rn.usn, rn.name, rs.status from v$rollname rn, v$rollstat rs where rn.usn = rs.usn;
USN NAME STATUS
---------- ------------------------------------------------------- ------------------------------
0 SYSTEM ONLINE
1 _SYSSMU1$ ONLINE
2 _SYSSMU2$ ONLINE
3 _SYSSMU3$ ONLINE
4 _SYSSMU4$ ONLINE
5 _SYSSMU5$ ONLINE
6 _SYSSMU6$ ONLINE
7 _SYSSMU7$ ONLINE
8 _SYSSMU8$ ONLINE
9 _SYSSMU9$ ONLINE
10 _SYSSMU10$ ONLINE
11 rows selected.
-- 仔細看，USN為0的回滾段為SYSTEM，這是我們創建數據庫的時候創建于系統表空間的回滾段，只提供于更改系統對象做服務。
sys@L10GR204> set line 150
sys@L10GR204> col SEGMENT_NAME for a15
sys@L10GR204> col TABLESPACE_NAME for a10
sys@L10GR204> select segment_name, segment_type, tablespace_name from dba_segments where segment_name = 'SYSTEM';
SEGMENT_NAME SEGMENT_TYPE TABLESPACE
--------------- ------------------------------------ ----------
SYSTEM ROLLBACK SYSTEM
如何證明SYSTEM回滾段只對系統對象提供服務
為了證明這句話的正確性，我先將非SYSTEM的回滾段去除，只要將當前UNDO表空間脫機處理即可。
sys@L10GR204> select file_id, tablespace_name from dba_data_files where tablespace_name = 'UNDOTBS1';
FILE_ID TABLESPACE
---------- ----------
2 UNDOTBS1
UNDO表空間脫機需要注意操作：
在非歸檔模式如下操作
OPEN模式直接操作：
sys@L10GR204> alter database datafile 2 offline drop;
ERROR:
ORA-03114: not connected to ORACLE
alter database datafile 2 offline drop
*
ERROR at line 1:
ORA-00603: ORACLE server session terminated by fatal error
告警日志頻繁輸出：
Mon Mar 25 13:35:28 2013
Errors in file /u01/app/oracle/admin/L10GR204/bdump/l10gr204_pmon_3756.trc:
ORA-00376: file 2 cannot be read at this time
ORA-01110: data file 2: '/u01/app/oracle/oradata/L10GR204/undotbs01.dbf'
Mon Mar 25 13:35:38 2013
Errors in file /u01/app/oracle/admin/L10GR204/bdump/l10gr204_pmon_3756.trc:
ORA-00376: file 2 cannot be read at this time
ORA-01110: data file 2: '/u01/app/oracle/oradata/L10GR204/undotbs01.dbf'
Mon Mar 25 13:35:38 2013
Errors in file /u01/app/oracle/admin/L10GR204/bdump/l10gr204_pmon_3756.trc:
ORA-00376: file 2 cannot be read at this time
ORA-01110: data file 2: '/u01/app/oracle/oradata/L10GR204/undotbs01.dbf'
Mon Mar 25 13:35:38 2013
Errors in file /u01/app/oracle/admin/L10GR204/bdump/l10gr204_pmon_3756.trc:
ORA-00376: file 2 cannot be read at this time
ORA-01110: data file 2: '/u01/app/oracle/oradata/L10GR204/undotbs01.dbf'
Mon Mar 25 13:35:38 2013
Errors in file /u01/app/oracle/admin/L10GR204/bdump/l10gr204_pmon_3756.trc:
ORA-00376: file 2 cannot be read at this time
ORA-01110: data file 2: '/u01/app/oracle/oradata/L10GR204/undotbs01.dbf'
試圖關閉數據庫
sys@L10GR204> shutdown immediate
-- 系統HANG住，無法正常關閉
解決方法：
sys@L10GR204> shutdown abort
sys@L10GR204> startup mount
sys@L10GR204> recover database;
sys@L10GR204> alter database open;
經過上面一系列操作，系統能open了，但是，當我們再次關閉數據庫的時候，將得到錯誤提示
sys@L10GR204> shutdown immediate
ORA-00376: file 2 cannot be read at this time
ORA-01110: data file 2: '/u01/app/oracle/oradata/L10GR204/undotbs01.dbf'
這是因為UNDO表空間當前狀態是OFFLINE緣故
sys@L10GR204> select FILE#, status from v$datafile;
FILE# STATUS
---------- --------------
1 SYSTEM
2 OFFLINE
3 ONLINE
4 ONLINE
5 ONLINE
在線處理之后恢復正常
sys@L10GR204> alter database datafile 2 online;
Database altered.
sys@L10GR204> select FILE#, status from v$datafile;
FILE# STATUS
---------- --------------
1 SYSTEM
2 ONLINE
3 ONLINE
4 ONLINE
5 ONLINE
sys@L10GR204> select rn.usn, rn.name, rs.status from v$rollname rn, v$rollstat rs where rn.usn = rs.usn;
USN NAME STATUS
---------- ------------------------------------------------------- ------------------------------
0 SYSTEM ONLINE
11 _SYSSMU11$ ONLINE
在數據庫open又開啟歸檔模式下，進行如下操作：
sys@L10GR204> alter database datafile 2 offline drop;
ERROR:
ORA-03114: not connected to ORACLE
alter database datafile 2 offline drop
*
ERROR at line 1:
ORA-00603: ORACLE server session terminated by fatal error
可見依然會掛起，但這時候，我們是無法通過非歸檔模式下解決方法來恢復，當打開數據庫時，直接異常終止，如下：
idle> alter database open;
alter database open
*
ERROR at line 1:
ORA-01092: ORACLE instance terminated. Disconnection forced
告警日志輸出：
Errors in file /u01/app/oracle/admin/L10GR204/bdump/l10gr204_smon_22399.trc:
ORA-00604: error occurred at recursive SQL level 1
ORA-00376: file 2 cannot be read at this time
ORA-01110: data file 2: '/u01/app/oracle/oradata/L10GR204/undotbs01.dbf'
Mon Mar 25 13:50:42 2013
Errors in file /u01/app/oracle/admin/L10GR204/bdump/l10gr204_smon_22399.trc:
ORA-00604: error occurred at recursive SQL level 1
ORA-00376: file 2 cannot be read at this time
ORA-01110: data file 2: '/u01/app/oracle/oradata/L10GR204/undotbs01.dbf'
Mon Mar 25 13:50:43 2013
Errors in file /u01/app/oracle/admin/L10GR204/udump/l10gr204_ora_23025.trc:
ORA-00604: error occurred at recursive SQL level 1
ORA-00376: file 2 cannot be read at this time
ORA-01110: data file 2: '/u01/app/oracle/oradata/L10GR204/undotbs01.dbf'
Error 604 happened during db open, shutting down database
USER: terminating instance due to error 604
Instance terminated by USER, pid = 23025
ORA-1092 signalled during: alter database open...
這時候，我們需要手工介恢復：
idle> conn / as sysdba
idle> startup mount
[oracle@khm21 ~]$ rman target /
RMAN> recover database;
idle> alter database open;
上面是在誤操作(OPEN狀態下)之后恢復的過程，OPEN狀態下將默認回滾表空間脫機處理是較危險的操作，請務必重啟到MOUNT模式下操作
sys@L10GR204> alter database datafile 2 offline drop;
-- 不帶drop選項，會報ORA-01145
sys@L10GR204> alter database datafile 2 offline;
alter database datafile 2 offline
*
ERROR at line 1:
ORA-01145: offline immediate disallowed unless media recovery enabled
在系統運行于歸檔模式下的時候，則如下操作：
sys@L10GR204> shutdown immediate
sys@L10GR204> startup mount
sys@L10GR204> alter database datafile 2 offline;
sys@L10GR204> alter database open;
下面驗證system回滾段只為系統對象做服務，使用sys用戶操作：
sys@L10GR204> create table test(id int);
Table created.
sys@L10GR204> insert into test values(1);
1 row created.
sys@L10GR204> delete from test;
1 row deleted.
sys@L10GR204> rollback;
Rollback complete.
當前只有一個SYSTEM回滾段，我們操作對象時SYSTEM表空間里的時候，SYSTEM回滾段就為此服務，因此正常。
sys@L10GR204> select rn.usn, rn.name, rs.status from v$rollname rn, v$rollstat rs where rn.usn = rs.usn;
USN NAME STATUS
---------- ------------------------------ ------------------------------
0 SYSTEM ONLINE
通過普通用戶操作：
ning@L10GR204> show user
USER is "ning"
ning@L10GR204> create table test(id int);
create table test(id int)
*
ERROR at line 1:
ORA-01552: cannot use system rollback segment for non-system tablespace 'ning'
-提示system回滾段無法為其他表空間提供服務。
將默認回滾表空間在線處理
sys@L10GR204> alter database datafile 2 online;
Database altered.
有了默認回滾段，操作恢復正常
ning@L10GR204> create table test(id int);
Table created.
ning@L10GR204> insert into test values(5);
1 row created.
ning@L10GR204> rollback;
Rollback complete.
這時候發現產生了一個回滾段
sys@L10GR204> select rn.usn, rn.name, rs.status from v$rollname rn, v$rollstat rs where rn.usn = rs.usn;
USN NAME STATUS
---------- ------------------------------ ------------------------------
0 SYSTEM ONLINE
1 _SYSSMU1$ ONLINE
通過如下方法可以查看_SYSSMU1$回滾段的使用情況：
sys@L10GR204> select OWNER, SEGMENT_NAME, TABLESPACE_NAME, EXTENT_ID, FILE_ID, BLOCK_ID, STATUS from dba_undo_extents where segment_name='_SYSSMU1$' and rownum <=3;
OWNER SEGMENT_NA TABLESPACE EXTENT_ID FILE_ID BLOCK_ID STATUS
------ ---------- ---------- ---------- ---------- ---------- ------------------
SYS _SYSSMU1$ UNDOTBS1 0 2 9 EXPIRED
SYS _SYSSMU1$ UNDOTBS1 1 2 2569 EXPIRED
SYS _SYSSMU1$ UNDOTBS1 2 2 24217 EXPIRED
alter system set undo_tablespace = uuu;
為什么要研究修改undo表空間，只是想告訴大家，當我們執行alter system set undo_tablespace = uuu;的時候都會發生什么。
首先我們觀察當前都有哪些回滾段：
sys@L10GR204> select rn.usn, rn.name, rs.status from v$rollname rn, v$rollstat rs where rn.usn = rs.usn;
USN NAME STATUS
---------- ------------------------------ ------------------------------
0 SYSTEM ONLINE
1 _SYSSMU1$ ONLINE
2 _SYSSMU2$ ONLINE
3 _SYSSMU3$ ONLINE
4 _SYSSMU4$ ONLINE
5 _SYSSMU5$ ONLINE
6 _SYSSMU6$ ONLINE
7 _SYSSMU7$ ONLINE
8 _SYSSMU8$ ONLINE
9 _SYSSMU9$ ONLINE
10 _SYSSMU10$ ONLINE
11 rows selected.
創建一個新的undo表空間，注意，這里為下面的實驗，我故意創建了非常小的回滾表空間。
sys@L10GR204> create undo tablespace undo_ning datafile '/u01/app/oracle/oradata/L10GR204/undo_ning.dbf' size 150K autoextend off;
Tablespace created.
將當前默認undo tablespace設置成新建的
sys@L10GR204> alter system set undo_tablespace=undo_ning;
System altered.
發現原先的回滾段全部消失，表空間過小，一個回滾段都沒創建。
sys@L10GR204> select rn.usn, rn.name, rs.status from v$rollname rn, v$rollstat rs where rn.usn = rs.usn;
USN NAME STATUS
---------- ------------------------------ ------------------------------
0 SYSTEM ONLINE
添加數據文件
sys@L10GR204> alter tablespace undo_ning add datafile '/u01/app/oracle/oradata/L10GR204/undo_ning02.dbf' size 200k autoextend on;
Tablespace altered.
生成一個回滾段。
sys@L10GR204> select rn.usn, rn.name, rs.status from v$rollname rn, v$rollstat rs where rn.usn = rs.usn;
USN NAME STATUS
---------- ------------------------------ ------------------------------
0 SYSTEM ONLINE
12 _SYSSMU12$ ONLINE
結論：在我們在線修改undo tablespace的時候，并不是原先的回滾段移動到新的undo tablespace，而是脫機重新創建。但如果原undo tablespace里還有active事務，則需要等到這些活動事務全部結束之后才能切換。
如何創建一個新的UNDO表空間并使用
上面已經演示過創建新UNDO表空間并使用的過程，這里分享一下如果創建過程遇到ORA-30015如何去解決。
sys@L10GR204> create undo tablespace l_u_s datafile '/u01/app/oracle/oradata/L10GR204/l_u_s01.dbf' size 1M autoextend off;
Tablespace created.
當我想使用新創建的undo表空間的時候，得到錯誤提示
sys@L10GR204> alter system set undo_tablespace=l_u_s;
alter system set undo_tablespace=l_u_s
*
ERROR at line 1:
ORA-02097: parameter cannot be modified because specified value is invalid
ORA-30015: previously offlined undo tablespace 'UNDO_ning' is still pending
查一下當前回滾段狀態，發現_SYSSMU19$狀態為PENDING OFFLINE，估計還有個活躍事務。
sys@L10GR204> col NAME for a20
sys@L10GR204> select rn.usn, rn.name, rs.status from v$rollname rn, v$rollstat rs where rn.usn = rs.usn;
USN NAME STATUS
---------- -------------------- ------------------------------
0 SYSTEM ONLINE
19 _SYSSMU19$ PENDING OFFLINE
34 _SYSSMU34$ ONLINE
35 _SYSSMU35$ ONLINE
36 _SYSSMU36$ ONLINE
37 _SYSSMU37$ ONLINE
38 _SYSSMU38$ ONLINE
39 _SYSSMU39$ ONLINE
40 _SYSSMU40$ ONLINE
9 rows selected.
我們通過下面腳本區找當前活躍事務SESSION信息
sys@L10GR204> col USERNAME for a10
sys@L10GR204> col ROLLBACK for a15
sys@L10GR204> SELECT S.SID, S.SERIAL#, S.USERNAME, R.NAME "ROLLBACK"
2 FROM V$SESSION S, V$TRANSACTION T, V$ROLLNAME R
3 WHERE S.TADDR = T.ADDR AND T.XIDUSN = R.USN
4 AND R.name IN (SELECT segment_name FROM dba_rollback_segs WHERE TABLESPACE_NAME ='UNDO_ning')
5 /
SID SERIAL# USERNAME ROLLBACK
---------- ---------- ---------- ---------------
148 34 ning _SYSSMU19$
直接將該SESSION KILL掉
sys@L10GR204> alter system kill session '148,34';
System altered.
這下就可以正常設置新UNDO表空間了
sys@L10GR204> alter system set undo_tablespace='L_U_S';
System altered.
sys@L10GR204> alter system set undo_retention=1;
System altered.
回滾段增減
在上面，我們已經看出，當Oracle數據庫啟動的時候，默認就會創建多個回滾段，但是，某些系統這些回滾段是不夠用的，表現在事務量非常大的系統中。
不過不用擔心，Oracle也知道有這種現象，因此如果當前回滾段不足，則會自動創建新回滾段并提供使用。一段時間(通過指定算法)后如果事務量減少，Oracle的SMON將部分回滾段脫機。
我們在上面環境的基礎下，通過實驗觀察這個效果。
這里用到了如下腳本
[oracle@khm21 ~]$ cat showsysevent.sql
column EVENT for a65
select *
from (
select event, total_waits, time_waited
from v$system_event
where wait_class <> 'Idle'
order by 3 desc
)
where EVENT = 'enq: US - contention'
/
查看當前回滾段情況：
sys@L10GR204> select rn.usn, rn.name, rs.status from v$rollname rn, v$rollstat rs where rn.usn = rs.usn;
USN NAME STATUS
---------- ------------------------------ ------------------------------
0 SYSTEM ONLINE
12 _SYSSMU12$ ONLINE
13 _SYSSMU13$ ONLINE
創建一張測試對象表并插入數據
ning@L10GR204> create table khm(id number, domain varchar2(20));
Table created.
ning@L10GR204> begin
2 for i in 1 .. 100 loop
3 insert into khm values(i,'Integrated Cloud Applications and Platform Services');
4
5 end loop;
6 commit;
7 end;
8 /
PL/SQL procedure successfully completed.
創建一個過程，對khm表進行delete，然后馬上回滾：
create or replace procedure khm_delete
is
begin
for i in 1 .. 100 loop
delete from khm;
rollback;
end loop;
end;
/
打開100個會話并執行該存儲過程
var jobno number;
begin
for l in 1 .. 100 loop
dbms_job.submit(:jobno, 'khm_delete;');
commit;
end loop;
end;
/
發現回滾段自動增加了幾個
sys@L10GR204> select rn.usn, rn.name, rs.status from v$rollname rn, v$rollstat rs where rn.usn = rs.usn;
USN NAME STATUS
---------- ------------------------------ ------------------------------
0 SYSTEM ONLINE
12 _SYSSMU12$ ONLINE
13 _SYSSMU13$ ONLINE
14 _SYSSMU14$ ONLINE
15 _SYSSMU15$ ONLINE
16 _SYSSMU16$ ONLINE
6 rows selected.
下面是US鎖等待統計，具體下面討論
ning@L10GR204> @showsysevent
EVENT TOTAL_WAITS TIME_WAITED
----------------------------------------------------------------- ----------- -----------
enq: US - contention 26 339
了解幾個參數
我不會去解釋所有和UNDO相關的參數，UNDO_MANAGEMENT、UNDO_RETENTION、UNDO_TABLESPACE這三個參數我也不解釋。
使用下面腳本查看和回滾相關的一些參數，了解一下即可，除非特殊情況不建議修改（尤其是隱含參數）
sys@L10GR204> set pagesize 9999
sys@L10GR204> set line 150
sys@L10GR204> col NAME for a30
sys@L10GR204> col VALUE for a20
sys@L10GR204> col DESCRIB for a80
sys@L10GR204> SELECT x.ksppinm NAME, y.ksppstvl VALUE, x.ksppdesc DESCRIB
2 FROM SYS.x$ksppi x, SYS.x$ksppcv y
3 WHERE x.inst_id = USERENV ('Instance')
4 AND y.inst_id = USERENV ('Instance')
5 AND x.indx = y.indx
6 AND x.ksppinm LIKE '%&par%'
7
sys@L10GR204> /
Enter value for par: undo
old 6: AND x.ksppinm LIKE '%&par%'
new 6: AND x.ksppinm LIKE '%undo%'
NAME VALUE DESCRIB
------------------------------ -------------------- --------------------------------------------------------------------------------
_gc_undo_affinity TRUE if TRUE, enable dynamic undo affinity
_kcl_undo_locks 128 number of locks per undo segment
_kcl_undo_grouping 32 grouping for undo block locks
_gc_undo_affinity_locks TRUE if TRUE, get affinity locks for undo
_gc_dissolve_undo_affinity FALSE if TRUE, dissolve undo affinity after an offline
_gc_initiate_undo_affinity TRUE if TRUE, initiate undo affinity after an online
undo_management AUTO instance runs in SMU mode if TRUE, else in RBU mode
undo_tablespace UNDO_ning use/switch undo tablespace
_collect_undo_stats TRUE Collect Statistics v$undostat
_undo_debug_mode 0 debug flag for undo related operations
_verify_undo_quota FALSE TRUE - verify consistency of undo quota statistics
_in_memory_undo TRUE Make in memory undo for top level transactions
_undo_autotune TRUE enable auto tuning of undo_retention
undo_retention 900 undo retention in seconds
_smon_undo_seg_rescan_limit 10 limit of SMON continous undo segments re-scan
_undo_debug_usage 0 invoke undo usage functions for testing
_optimizer_undo_cost_change 10.2.0.4 optimizer undo cost change
_optimizer_undo_changes FALSE undo changes to query optimizer
18 rows selected.
sys@L10GR204> SELECT x.ksppinm NAME, y.ksppstvl VALUE, x.ksppdesc DESCRIB
2 FROM SYS.x$ksppi x, SYS.x$ksppcv y
3 WHERE x.inst_id = USERENV ('Instance')
4 AND y.inst_id = USERENV ('Instance')
5 AND x.indx = y.indx
6 AND x.ksppinm LIKE '%&par%'
7
sys@L10GR204> /
Enter value for par: rollback
old 6: AND x.ksppinm LIKE '%&par%'
new 6: AND x.ksppinm LIKE '%rollback%'
NAME VALUE DESCRIB
------------------------------ -------------------- --------------------------------------------------------------------------------
transactions_per_rollback_segm 5 number of active transactions per rollback segment
ent
rollback_segments undo segment list
_rollback_segment_initial 1 starting undo segment number
_rollback_segment_count 0 number of undo segments
_offline_rollback_segments offline undo segment list
_corrupted_rollback_segments corrupted undo segment list
_cleanup_rollback_entries 100 no. of undo entries to apply per transaction cleanup
_rollback_stopat 0 stop at -position to step rollback
fast_start_parallel_rollback LOW max number of parallel recovery slaves that may be used
9 rows selected.
從中介紹幾個隱含參數：
_undo_autotune ：從Oracle 10g R2開始默認為TRUE，意思是根據undo表空間大小或者系統負載情況，自動調整undo_retention。這個功能可以在線關閉，方法：alter system set "_undo_autotune"=false;
_smon_undo_seg_rescan_limit：limit of SMON continous undo segments re-scan，上面介紹了Oracle脫機回滾段是根據算法的，算法里面的其中一個條件（如果前面條件都不滿足則采用此條件）就是參考_smon_undo_seg_rescan_limit的值。
_rollback_segment_count：指定在線狀態的回滾段個數。
構造一個CR塊
下面是簡單構造一個CR塊的示例。
SESSION A:
ning@L10GR204> select distinct sid from v$mystat;
SID
----------
149
當前KHM表id=95的記錄如下：
ning@L10GR204> select * from khm where id = 95;
ID DOMAIN
---------- ----------------------------------------
95 Integrated Cloud Applications and Platform Services
打開一個游標，但不執行
ning@L10GR204> variable l refcursor
ning@L10GR204> exec open :l for select * from khm where id = 95;
PL/SQL procedure successfully completed.
SESSION B:
ning@L10GR204> select distinct sid from v$mystat;
SID
----------
143
在另一個會話將id=95的記錄值修改并提交
ning@L10GR204> update khm set domain = 'Create your own social network with the best community website builder - NING' where id = 95;
1 row updated.
ning@L10GR204> commit;
Commit complete.
回到打開游標的會話，執行并讀取，這時我們發現讀取到的是被SESSION B修改之前的記錄，這就從CR塊里讀出來的記錄
SESSION A:
ning@L10GR204> print l
ID DOMAIN
---------- ----------------------------------------
95 Integrated Cloud Applications and Platform Services
再次查詢，因為事務已經提交，所以讀取到的是UPDATE之后的內容
ning@L10GR204> select * from khm where id = 95;
ID DOMAIN
---------- ----------------------------------------
95 Create your own social network with the best community website builder - NING
這其中SCN的作用是非常重要的，其實SESSION A打開游標的時候就有對應的SCN，這個SCN肯定小于更新并提交時SCN，當我們print的時候還是拿當前SCN去查詢，經過跟塊上的SCN比較之后它去回滾段讀取鏡像數據。
下面我們深入了解一下構造CR塊并讀取過程。
拿下面數據作為參考進行實驗：
ning@L10GR204> select rowid, t.* from khm t where t.id >95;
ROWID ID DOMAIN
------------------ ---------- ----------------------------------------
AAANIuAAFAAAAAWABf 96 Create your own social network with the best community website builder - NING
AAANIuAAFAAAAAWABg 97 Create your own social network with the best community website builder - NING
AAANIuAAFAAAAAWABh 98 Create your own social network with the best community website builder - NING
AAANIuAAFAAAAAWABi 99 Integrated Cloud Applications and Platform Services
AAANIuAAFAAAAAWABj 100 Integrated Cloud Applications and Platform Services
SESSION A更新兩條記錄
ning@L10GR204> update khm set domain='bbs.ning.com' where id=96;
1 row updated.
ning@L10GR204> update khm set domain='ffdg' where id=97;
1 row updated.
查看該事務存在信息
ning@L10GR204> @showtra
Enter value for sid: 149
old 13: WHERE SID=&SID
new 13: WHERE SID=149
XIDUSN XIDSLOT XIDSQN UBAFIL UBABLK UBASQN UBAREC STATUS
---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------------
50 12 3 13 44 3 45 ACTIVE
通過USN號去查看當前使用的回滾段
sys@L10GR204> select USN, NAME from v$rollname where USN=50;
USN NAME
---------- -------------------------
50 _SYSSMU50$
如果想dump該回滾段頭信息，則如下方式操作：
sys@L10GR204> alter system dump undo header '_SYSSMU50$';
sys@L10GR204> oradebug setmypid
sys@L10GR204> oradebug tracefile_name
/u01/app/oracle/admin/L10GR204/udump/l10gr204_ora_4085.trc
如果想dump該回滾段內容看，則如下方式操作：
sys@L10GR204> alter system dump datafile 13 block 44;
sys@L10GR204> oradebug setmypid
sys@L10GR204> oradebug tracefile_name
/u01/app/oracle/admin/L10GR204/udump/l10gr204_ora_4277.trc
至于DUMP出來的內容如何去看，我將在下面分析，這里不做解釋。
下面是簡單演示如何通過v$bh查看cr狀態xcur狀態：
SESSION A:
當前是free狀態
sys@L10GR204> select file#, block#, status from v$bh where file#=5 and block#=23;
FILE# BLOCK# STATUS
---------- ---------- --------------
5 23 free
插入一條記錄并查看
ning@L10GR204> insert into khm values(1,'恩墨學院 - 名師傳道，授業以專');
1 row created.
ning@L10GR204> @showrowid
Enter value for table_name: khm
old 7: from &table_name
new 7: from khm
Enter value for id: 1
old 8: where id=&id
new 8: where id=1
ROWID OBJECT_ID FILE_ID BLOCK_ID NUM ROWIDTOCHAR(ROWID)
------------------ ---------- ---------- ---------- ---------- ------------------------------------
AAANIuAAFAAAAAXAAA 53806 5 23 0 AAANIuAAFAAAAAXAAA
sys@L10GR204> select file#, block#, status from v$bh where file#=5 and block#=23;
FILE# BLOCK# STATUS
---------- ---------- --------------
5 23 free
5 23 xcur
5 23 cr
-- INSERT一條記錄產生了一個XCUR塊兒以及CR塊兒。
SESSION B:
查詢該表：
ning@L10GR204> select * from khm;
ID DOMAIN
---------- ----------------------------------------
91 Create your own social network with the best community website builder - NING
92 Create your own social network with the best community website builder - NING
93 Create your own social network with the best community website builder - NING
94 Create your own social network with the best community website builder - NING
95 Create your own social network with the best community website builder - NING
96 Create your own social network with the best community website builder - NING
97 Create your own social network with the best community website builder - NING
98 Create your own social network with the best community website builder - NING
99 Integrated Cloud Applications and Platform Services
100 Integrated Cloud Applications and Platform Services
10 rows selected.
結果又產生了一個CR塊兒
sys@L10GR204> select file#, block#, status from v$bh where file#=5 and block#=23;
FILE# BLOCK# STATUS
---------- ---------- --------------
5 23 free
5 23 xcur
5 23 cr
5 23 cr
SESSION C:
ning@L10GR204> select * from khm;
ID DOMAIN
---------- ----------------------------------------
91 Create your own social network with the best community website builder - NING
92 Create your own social network with the best community website builder - NING
93 Create your own social network with the best community website builder - NING
94 Create your own social network with the best community website builder - NING
95 Create your own social network with the best community website builder - NING
96 Create your own social network with the best community website builder - NING
97 Create your own social network with the best community website builder - NING
98 Create your own social network with the best community website builder - NING
99 Integrated Cloud Applications and Platform Services
100 Integrated Cloud Applications and Platform Services
10 rows selected.
sys@L10GR204> select file#, block#, status from v$bh where file#=5 and block#=23;
FILE# BLOCK# STATUS
---------- ---------- --------------
5 23 free
5 23 cr
5 23 cr
5 23 xcur
5 23 cr
SESSION D、SESSION E、SESSION F...也同樣操作，結果是CR塊兒最多構造5個，超過5個則覆蓋之前的CR塊兒。
sys@L10GR204> select file#, block#, status from v$bh where file#=5 and block#=23;
FILE# BLOCK# STATUS
---------- ---------- --------------
5 23 free
5 23 cr
5 23 cr
5 23 cr
5 23 cr
5 23 xcur
5 23 cr
7 rows selected.
CR最多創建多少個受隱含參數_db_block_max_cr_dba值的影響，默認為6。
sys@L10GR204> @GetHparDes
Enter value for par: block_max_cr
old 6: AND x.ksppinm LIKE '%&par%'
new 6: AND x.ksppinm LIKE '%block_max_cr%'
NAME VALUE DESCRIB
------------------------------ -------------------- --------------------------------------------------------------------------------
_db_block_max_cr_dba 6 Maximum Allowed Number of CR buffers per dba
快速構造ora-01555 快照過舊
ORA-01555是一個ORACLE非常經典的錯誤之一，研究UNDO的時候不能排除的內容。
ning@L10GR204> host oerr ora 1555
01555, 00000, "snapshot too old: rollback segment number %s with name "%s" too small"
// *Cause: rollback records needed by a reader for consistent read are
// overwritten by other writers
// *Action: If in Automatic Undo Management mode, increase undo_retention
// setting. Otherwise, use larger rollback segments
查看當前UNDO相關參數設置以及undo表空間大小
sys@L10GR204> show parameter undo
NAME TYPE VALUE
------------------------------------ ---------------------- ------------------------------
undo_management string AUTO
undo_retention integer 2
undo_tablespace string L_S_U
sys@L10GR204> select bytes/1048576, AUTOEXTENSIBLE from dba_data_files where tablespace_name='L_S_U';
BYTES/1048576 AUTOEX
------------- ------
1 NO
創建一個過程：
create or replace procedure ltb_update
is
cursor c1 is select object_id from ltb where rownum <= 100;
begin
for r_c1 in c1 loop
update ltb set owner='ning' where object_id=r_c1.object_id;
commit;
end loop;
end;
/
打開30個會話并執行該存儲過程
var jobno number;
begin
for l in 1 .. 20 loop
dbms_job.submit(:jobno, 'ltb_update;');
commit;
end loop;
end;
/
告警日志輸出大量的ORA-015555：
Tue Mar 26 21:59:21 2013
ORA-01555 caused by SQL statement below (SQL ID: a9s2m307s4dv1, Query Duration=1 sec, SCN: 0x0000.0046a23b):
Tue Mar 26 21:59:21 2013
UPDATE LTB SET OWNER='ning' WHERE OBJECT_ID=:B1
Tue Mar 26 21:59:21 2013
ORA-01555 caused by SQL statement below (SQL ID: a9s2m307s4dv1, Query Duration=1 sec, SCN: 0x0000.0046a1df):
Tue Mar 26 21:59:21 2013
UPDATE LTB SET OWNER='ning' WHERE OBJECT_ID=:B1
Tue Mar 26 21:59:21 2013
Errors in file /u01/app/oracle/admin/L10GR204/bdump/l10gr204_j020_5371.trc:
ORA-12012: error on auto execute of job 2153
ORA-01555: snapshot too old: rollback segment number 51 with name "_SYSSMU51$" too small
ORA-06512: at "ning.LTB_UPDATE", line 6
ORA-06512: at line 1
Tue Mar 26 21:59:21 2013
Errors in file /u01/app/oracle/admin/L10GR204/bdump/l10gr204_j012_5354.trc:
ORA-12012: error on auto execute of job 2126
ORA-01555: snapshot too old: rollback segment number 54 with name "_SYSSMU54$" too small
ORA-06512: at "ning.LTB_UPDATE", line 6
ORA-06512: at line 1
最后討論下ORA-01555的解決方案：
1） 不要讓事務很大；
2） 允許undo表空間自動擴展，undo_retention根據需求設大些；
3） 減少不必要的commit；
4） 優化SQL，減少查詢時間。
enq: US – contention
上面經過測試，在事務數增加導致回滾段不足時，Oracle會分配更多回滾段，這時候服務器進程或者后臺進程需要獲得US鎖，每個回滾段都分配一個US鎖，如果這個獲取過程中發生爭用，就會產生enq: US – contention等待事件。一般情況下，US鎖爭用現象很少發生，但如果發生了，需要具體分析，從應用角度分析，或者查看BUG信息。
在上面我提到過，Oracle的SMON進程將部分回滾段脫機處理，肯定有朋友不清楚SMON居然還干這事。這是Oracle為了減少撤銷空間使用的壓力而設計的。
但是SMON并不是經常做這件事情，這里有個算法，SMON基于過去7天v$undostat動態視圖信息或者AWR中UNDO歷史快照使用信息來決定脫機回滾段數量，而且不刪除多余的回滾段，而是OFFLINE處理。
如果發現脫機回滾段嚴重影響性能，DBA也可以禁止此功能，方法：10511事件。
sys@L10GR204> host oerr ora 10511
10511, 00000, "turn off SMON check to cleanup undo dictionary"
// *Cause:
// *Action:
sys@L10GR204> alter system set events '10511 trace name context forever, level 1';
System altered.
SMON還為UNDO段做一件事情，就是定期收縮(Shrink)。當然只有SMON才定期做這件事情，不一定SMON不工作回滾段就不收縮，比如回滾段之間空間緊張時也會發生。
SMON每12個小時進行一次收縮，這個功能我們也可以使用events事件禁用，但一般不建議這么做。
sys@L10GR204> host oerr ora 10512
10512, 00000, "turn off SMON check to shrink rollback segments"
// *Cause:
// *Action:
sys@L10GR204> alter system set events '10512 trace name context forever,level 1';
System altered.
Permalink
Leave a Comment
姓名 (必填)
電子郵件 required)
URI
所見即所得編輯器
工具欄 源碼 插入代碼
▲
元素路徑
NOTICE: You should type some Chinese word (like “你好”) in your comment to pass the spam-check, thanks for your patience!
來自 <http://www.ning.com/archives/755.html>
總結
以上是生活随笔為你收集整理的oracle segment undo_71_UNDO扩展学习的全部內容，希望文章能夠幫你解決所遇到的問題。
- 上一篇： 【Python】Pygame模块设计游戏
- 下一篇： 移动端iphone抓包调试神器—Stre