验证ADG的坏块检测和自动修复

1.主库新建测试文件

主库在AWR的PDB中做测试，为了不影响其他测试，创建一个新的测试表空间tbs_test及对应数据文件：

 

SQL> conn awr@awr
Enter password:
Connected.
SQL> create tablespace tbs_test datafile '/flash/oradata/DEMO/awr/tbs_test01.dbf' size 30m;

Tablespace created.

2.主库创建测试表

主库在新建表空间上创建测试表awr.test：

SQL> create table awr.test tablespace tbs_test as select * from dba_users;

Table created.

SQL> select count(*) from awr.test;

  COUNT(*)
----------
	37

3.查询表对应数据文件信息

通过dbms_rowid查看awr.test表对应行数据的文件号（rel_fno）、块号（blockno,）和行号（rowno）:

select rowid, 
     dbms_rowid.rowid_relative_fno(rowid) rel_fno,        
     dbms_rowid.rowid_block_number(rowid) blockno,  
     dbms_rowid.rowid_row_number(rowid) rowno 
from awr.test    
order by rowid;

ROWID		      REL_FNO	 BLOCKNO      ROWNO
------------------ ---------- ---------- ----------
AAATR2AAdAAAACDAAA	   29	     131	  0
AAATR2AAdAAAACDAAB	   29	     131	  1
AAATR2AAdAAAACDAAC	   29	     131	  2
AAATR2AAdAAAACDAAD	   29	     131	  3
AAATR2AAdAAAACDAAE	   29	     131	  4
AAATR2AAdAAAACDAAF	   29	     131	  5
AAATR2AAdAAAACDAAG	   29	     131	  6
AAATR2AAdAAAACDAAH	   29	     131	  7
AAATR2AAdAAAACDAAI	   29	     131	  8
AAATR2AAdAAAACDAAJ	   29	     131	  9
AAATR2AAdAAAACDAAK	   29	     131	 10
...

4.模拟数据文件物理坏块

使用dd模拟数据文件的物理坏块：

dd if=/dev/zero of=/flash/oradata/DEMO/awr/tbs_test01.dbf bs=8192 conv=notrunc seek=131 count=1

5.查询对应测试表

再次查询被破坏数据文件上的表awr.test，发现客户端只是卡顿一下就正常出了结果，并没有任何显示的报错：

ALTER SYSTEM Flush buffer_cache;
select count(*) from awr.test;

6.进一步查询日志信息

上面查询表没有报错，但是从主库的alert日志中可以看到：

2023-04-03T12:08:02.602504+08:00
AWR(6):create tablespace tbs_test datafile '/flash/oradata/DEMO/awr/tbs_test01.dbf' size 30m
AWR(6):Completed: create tablespace tbs_test datafile '/flash/oradata/DEMO/awr/tbs_test01.dbf' size 30m
2023-04-03T12:15:21.021834+08:00
AWR(6):ALTER SYSTEM: Flushing buffer cache inst=0 container=6 global
AWR(6):TABLE AUDSYS.AUD$UNIFIED: ADDED INTERVAL PARTITION SYS_P368 (106) VALUES LESS THAN (TIMESTAMP' 2023-05-01 00:00:00')
2023-04-03T12:15:24.751443+08:00
AWR(6):Hex dump of (file 29, block 131) in trace file /u01/app/oracle/diag/rdbms/demo/demo/trace/demo_ora_11735.trc
AWR(6):
AWR(6):Corrupt block relative dba: 0x07400083 (file 29, block 131)
AWR(6):Completely zero block found during multiblock buffer read
AWR(6):
AWR(6):Reading datafile '/flash/oradata/DEMO/awr/tbs_test01.dbf' for corrupt data at rdba: 0x07400083 (file 29, block 131)
AWR(6):Reread (file 29, block 131) found same corrupt data (no logical check)
AWR(6):Starting background process ABMR
2023-04-03T12:15:24.763408+08:00
Corrupt Block Found
         TIME STAMP (GMT) = 04/03/2023 12:15:24
         CONT = 6, TSN = 5, TSNAME = TBS_TEST
         RFN = 29, BLK = 131, RDBA = 121634947
         OBJN = 78966, OBJD = 78966, OBJECT = TEST, SUBOBJECT =
         SEGMENT OWNER = AWR, SEGMENT TYPE = Table Segment
2023-04-03T12:15:24.766521+08:00
ABMR started with pid=132, OS id=11983
2023-04-03T12:15:24.767751+08:00
Automatic block media recovery service is active.
2023-04-03T12:15:24.767981+08:00
AWR(6):Automatic block media recovery requested for (file# 29, block# 131)
2023-04-03T12:15:27.096763+08:00
Automatic block media recovery successful for (file# 29, block# 131)
2023-04-03T12:15:27.097189+08:00
AWR(6):Automatic block media recovery successful for (file# 29, block# 131)

日志中显示自动启用了ABMR（Automatic block media recovery）成功修复了物理坏块。

7.确认当前参数设置

如果查询 db_block_checking 、 db_lost_write_protect 这些参数，会发现我这里并没有去特殊设置：

SQL> show parameter db_block

NAME				     TYPE	 VALUE
------------------------------------ ----------- ------------------------------
db_block_buffers		     integer	 0
db_block_checking		     string	 FALSE
db_block_checksum		     string	 TYPICAL
db_block_size			     integer	 8192
SQL> show parameter db_lost

NAME				     TYPE	 VALUE
------------------------------------ ----------- ------------------------------
db_lost_write_protect		     string	 NONE
SQL>

那么那些参数的意义呢？其实MOS文档：

• Best Practices for Corruption Detection, Prevention, and Automatic Repair - in a Data Guard Configuration (Doc ID 1302539.1)

文档中有说明，物理坏块默认ADG就能检测，逻辑坏块要配合这些参数设置。包括上一步的日志信息中，在发现数据损坏时，也标注了(no logical check)非逻辑检查的提示。

当然，如果您想要获得更全面的保护，还是要按文档说明，额外设置这些参数。