13 oracle數據庫壞塊-邏輯壞塊
邏輯數據壞塊的場景
1）oracle bug也可能導致邏輯壞塊的產生. 特別是parallel dml. 例如:
Bug 5621677 Logical corruption with PARALLEL update
Bug 6994194 Logical corruption from UPDATE DML
Bug 15980234 ORA-1400 / logical corruption from direct path INSERT ALL(fail with
ORA-1400, in direct path / PDML) 。
2）多數情況下邏輯壞塊可能都是軟件問題導致，當然數據庫異常也可能導致。比如掉電的情況下，
就可能導致塊內數據不一致
邏輯數據壞塊檢測工具
RMAN備份恢忽略soft corruption
soft corruption的塊不計入maxcorrupt
media recovery會忽略soft corrupt
RMAN validate命令不會在alert中記錄soft corrupt的信息，但是會在v$database_block_corruption中記錄
DBV可以檢測soft corruption
如果不設置event10231或者類似事件，那么soft corruption的塊再次訪問時報ORA-1578
邏輯數據壞塊的檢測
exp\expdp 該方式存在缺陷,是發現邏輯壞塊后就會報錯,不再繼續檢測.如果是索引邏輯壞塊,那么exp是檢測不到的
rman 同檢測物理壞塊一樣,rman備份也具備檢測邏輯壞塊的功能，而且用的很多.
通常我們在進行數據庫不完全恢復時，很多時候可能需要使用隱含參數強制打開數據庫,打開之后我們都建議
使用rman進行檢查，確認是否存在邏輯壞塊等等,以免open后數據庫出現異常
bbed 上面有介紹
sql查詢 select select count(1) table_name ;
select /*+index_ffs(xxx) */ count(1) from table_name;
analyze命令 analyze index idx_name validate structure;
analyze table idx_name validate structure cascade online;
db_block_checking參數 在10g版本中，該參數默認是false,如果設置為true,那么可以進行邏輯壞塊的檢測,實際上設置以后, oracle在寫入時就會進行check，避免產生邏輯壞塊
邏輯壞塊的分類
data block 通常我們是最大程度的將segment內的數據搶救出來，然后處理掉壞塊.(如果是個別壞塊,可以手工修復)
index block 對於index logical 壞塊,一般來講,我們通過rebuild就可以解決(有時或許不行,可以drop然后重建)
segment header/bitmap block ...etc 這種情況下比較復雜
模擬邏輯壞塊
在很多情況下邏輯壞塊都發生在索引上,所以這里我們以模擬索引邏輯壞塊為例進行說明
YHQT@ orcl >create table yhqtest_6 as select owner,object_id,object_name from dba_objects where rownum<300;
Table created.
YHQT@ orcl >create index idx_yhqt6_id on yhqtest_6(object_id);
Index created.
YHQT@ orcl >select owner,object_id from dba_objects where object_name=upper('idx_yhqt6_id');
OWNER OBJECT_ID
------------------------------ ----------
YHQT 88407
YHQT@ orcl >alter session set events 'immediate trace name treedump level 88407';
Session altered.
--查看trace結果
[oracle@DSI ~]$ more /u01/app/oracle/diag/rdbms/orcl/orcl/trace/orcl_ora_2354.trc
*** 2019-07-15 16:00:53.250
*** MODULE NAME:(SQL*Plus) 2019-07-15 16:00:53.250
*** ACTION NAME:() 2019-07-15 16:00:53.250
----- begin tree dump
leaf: 0x28000db 41943259 (0: nrow: 299 rrow: 299)
----- end tree dump
YHQT@ orcl >select dbms_utility.data_block_address_file(TO_NUMBER('28000db','XXXXXXXX')) file_id,
dbms_utility.data_block_address_block(TO_NUMBER('28000db','XXXXXXXX')) block_id from dual; 2
FILE_ID BLOCK_ID
---------- ----------
10 219
--使用bbed查看並修改
BBED> set file 10 block 219
FILE# 10
BLOCK# 219
BBED> map /v
File: /u01/app/oracle/oradata/orcl/yhqt01.dbf (10)
Block: 219 Dba:0x028000db
------------------------------------------------------------
BBED> p kdxle
struct kdxle, 32 bytes @100
struct kdxlexco, 16 bytes @100
ub1 kdxcolev @100 0x00
ub1 kdxcolok @101 0x00
ub1 kdxcoopc @102 0x80
ub1 kdxconco @103 0x02
ub4 kdxcosdc @104 0x00000000
sb2 kdxconro @108 299
sb2 kdxcofbo @110 634
sb2 kdxcofeo @112 4246
sb2 kdxcoavs @114 3612
sb2 kdxlespl @116 0
sb2 kdxlende @118 0
ub4 kdxlenxt @120 0x00000000
ub4 kdxleprv @124 0x00000000
ub1 kdxledsz @128 0x00
ub1 kdxleflg @129 0x00 (NONE)
BBED> dump /v offset 108 count 32
File: /u01/app/oracle/oradata/orcl/yhqt01.dbf (10)
Block: 219 Offsets: 108 to 139 Dba:0x028000db
-------------------------------------------------------
2b017a02 96101c0e 00000000 00000000 l +.z.............
YHQT@ orcl >select to_char(299,'xxxxxx') from dual;
TO_CHAR
-------
12b
BBED> modify /x 2a offset 108
File: /u01/app/oracle/oradata/orcl/yhqt01.dbf (10)
Block: 219 Offsets: 108 to 139 Dba:0x028000db
------------------------------------------------------------------------
2a017a02 96101c0e 00000000 00000000 00000000 00000000 601f0000 541f481f
<32 bytes per line>
BBED> sum apply
Check value for File 10, Block 219:
current = 0xf239, required = 0xf239
使用rman進行查看
RMAN> backup validate check logical database;
File Status Marked Corrupt Empty Blocks Blocks Examined High SCN
---- ------ -------------- ------------ --------------- ----------
5 FAILED 0 81 64011 10373342
File Name: /home/oracle/backup/test01.tts
Block Type Blocks Failing Blocks Processed
---------- -------------- ----------------
Data 3 20628
Index 0 13088
Other 0 30203
File Status Marked Corrupt Empty Blocks Blocks Examined High SCN
---- ------ -------------- ------------ --------------- ----------
10 FAILED 0 6190 6400 10411019
File Name: /u01/app/oracle/oradata/orcl/yhqt01.dbf
Block Type Blocks Failing Blocks Processed
---------- -------------- ----------------
Data 2 53
Index 1 2
Other 1 155
用視圖查詢
YHQT@ orcl >select * from v$database_block_corruption;
FILE# BLOCK# BLOCKS CORRUPTION_CHANGE# CORRUPTIO
---------- ---------- ---------- ------------------ ---------
10 135 1 0 FRACTURED
10 160 1 0 ALL ZERO
5 2199 1 9533291 CORRUPT
5 2207 1 9579846 CORRUPT
5 2231 1 9620379 CORRUPT
10 151 1 10410172 CORRUPT
10 219 1 10411020 CORRUPT
CORRUPTION_CHANGE#值= 10410172,我們可以發現,corruption_change#值是大於0,那么表示是邏輯壞塊
--查詢壞塊所在的對象
set autot off set lines 150 col segment_name for a15 col owner for a20 SELECT tablespace_name, segment_type, owner, segment_name FROM dba_extents WHERE file_id = 10 and 219 between block_id AND block_id + blocks - 1;
YHQT@ orcl >SELECT tablespace_name, segment_type, owner, segment_name FROM dba_extents WHERE file_id = 10 and 219 between block_id AND block_id + blocks - 1; 2 3 4 TABLESPACE_NAME SEGMENT_TYPE OWNER SEGMENT_NAME ------------------------------ ------------------ -------------------- --------------- YHQT INDEX YHQT IDX_YHQT6_ID
使用dbms_repair檢測
創建repair table YHQT@ orcl >conn / as sysdba YHQT@ orcl >BEGIN DBMS_REPAIR.ADMIN_TABLES ( TABLE_NAME => 'REPAIR_TABLE', TABLE_TYPE => dbms_repair.repair_table, ACTION => dbms_repair.create_action, TABLESPACE => 'YHQT'); END; /
檢查對象上是否存在壞塊 YHQT@ orcl >set serveroutput on DECLARE num_corrupt INT; BEGIN num_corrupt := 0; DBMS_REPAIR.CHECK_OBJECT ( SCHEMA_NAME => 'YHQT', OBJECT_NAME => 'YHQTEST_6', REPAIR_TABLE_NAME => 'REPAIR_TABLE', corrupt_count => num_corrupt); DBMS_OUTPUT.PUT_LINE('number corrupt: ' || TO_CHAR (num_corrupt)); END; / 通過check后,就可以查詢repair table了 SYS@ orcl >col CORRUPT_DESCRIPTION for a100 SYS@ orcl >select OBJECT_ID,RELATIVE_FILE_ID,BLOCK_ID,CORRUPT_TYPE,OBJECT_NAME,CORRUPT_DESCRIPTION FROM repair_table;
---修復邏輯塊錯誤--index
SYS@ orcl >conn yhqt/*** Connected. YHQT@ orcl >alter index IDX_YHQT6_ID rebuild; Index altered. YHQT@ orcl >select * from v$database_block_corruption; FILE# BLOCK# BLOCKS CORRUPTION_CHANGE# CORRUPTIO ---------- ---------- ---------- ------------------ --------- 10 135 1 0 FRACTURED 10 160 1 0 ALL ZERO 5 2199 1 9533291 CORRUPT 5 2207 1 9579846 CORRUPT 5 2231 1 9620379 CORRUPT 10 151 1 10410172 CORRUPT 10 219 1 10411020 CORRUPT 7 rows selected. YHQT@ orcl >analyze table yhqtest_6 validate structure cascade online; analyze table yhqtest_6 validate structure cascade online * ERROR at line 1: ORA-01499: table/index cross reference failure - see trace file YHQT@ orcl >drop index IDX_YHQT6_ID; Index dropped. YHQT@ orcl >create index idx_yhqt6_id on yhqtest_6(object_id); Index created. YHQT@ orcl >analyze table yhqtest_6 validate structure cascade online; Table analyzed.
使用expdp備份
[oracle@DSI ~]$ expdp system/**** DIRECTORY=dump_file_dir_test compression=all schemas=YHQT dumpfile=schemas_YHQT_%U.DMP parallel=2 logfile=schemas_YHQT_20190715.log
修復上一篇物理壞塊測試中的
2.The block is Fractured/Incomplete - header and footer of the block do not match
分析alert日志
Mon Jul 15 11:12:19 2019 Hex dump of (file 10, block 143) in trace file /u01/app/oracle/diag/rdbms/orcl/orcl/trace/orcl_ora_2354.trc Corrupt block relative dba: 0x0280008f (file 10, block 143) Fractured block found during multiblock buffer read Data in bad block: type: 6 format: 2 rdba: 0x0280008f last change scn: 0x0000.009e53f5 seq: 0x1 flg: 0x06 spare1: 0x0 spare2: 0x0 spare3: 0x0 consistency value in tail: 0x54f50601 ===>>>tail=2 lower bytes of SCN Base+type+seq=53f5+06+01 而改變之后的值是54f50601 check value in block header: 0x46ca computed block checksum: 0x0 Reading datafile '/u01/app/oracle/oradata/orcl/yhqt01.dbf' for corruption at rdba: 0x0280008f (file 10, block 143) Reread (file 10, block 143) found same corrupt data (no logical check) Mon Jul 15 11:12:19 2019 Corrupt Block Found TSN = 10, TSNAME = YHQT RFN = 10, BLK = 143, RDBA = 41943183 OBJN = 88385, OBJD = 88385, OBJECT = YHQTEST_2, SUBOBJECT = SEGMENT OWNER = YHQT, SEGMENT TYPE = Table Segment Corrupt Block Found TSN = 10, TSNAME = YHQT RFN = 10, BLK = 143, RDBA = 41943183 OBJN = 88385, OBJD = 88385, OBJECT = YHQTEST_2, SUBOBJECT = SEGMENT OWNER = YHQT, SEGMENT TYPE = Table Segment Errors in file /u01/app/oracle/diag/rdbms/orcl/orcl/trace/orcl_ora_2354.trc (incident=67366): ORA-01578: ORACLE data block corrupted (file # 10, block # 143) ORA-01110: data file 10: '/u01/app/oracle/oradata/orcl/yhqt01.dbf' Incident details in: /u01/app/oracle/diag/rdbms/orcl/orcl/incident/incdir_67366/orcl_ora_2354_i67366.trc Mon Jul 15 11:12:21 2019 Sweep [inc][67366]: completed Errors in file /u01/app/oracle/diag/rdbms/orcl/orcl/trace/orcl_ora_2354.trc (incident=67367): ORA-01578: ORACLE data block corrupted (file # 10, block # 143) ORA-01110: data file 10: '/u01/app/oracle/oradata/orcl/yhqt01.dbf' Incident details in: /u01/app/oracle/diag/rdbms/orcl/orcl/incident/incdir_67367/orcl_ora_2354_i67367.trc Hex dump of (file 10, block 135) in trace file /u01/app/oracle/diag/rdbms/orcl/orcl/incident/incdir_67366/orcl_m000_2539_i67366_a.trc
用bbed修改
BBED> p tailchk ub4 tailchk @8188 0x54f50601 BBED> modify /x 0106f553 offset 8188 Warning: contents of previous BIFILE will be lost. Proceed? (Y/N) y File: /u01/app/oracle/oradata/orcl/yhqt01.dbf (10) Block: 143 Offsets: 8188 to 8191 Dba:0x0280008f ------------------------------------------------------------------------ 0106f553 <32 bytes per line> BBED> sum apply Check value for File 10, Block 143: current = 0x41ca, required = 0x41ca
YHQT@ orcl >alter system flush buffer_cache; YHQT@ orcl >set linesize 999 YHQT@ orcl >select * from yhqtest_2; ID NAME ---------- ---------===查詢正常，數據也還在 1 yhq