12c pdb下的 ABMR测试 Automatic Block Media Recovery by Active Data Guard

- Automatic Block Media Recovery by Active Data Guard

ABMR 特性从 11g 版本开始

ABMR 无需额外设置 db_lost_write_protect参数.

ABMR 默认打开 "_auto_bmr" (默认为enabled)

[oracle@wh ~]$ ./orz.sh param | grep db_lost

db_lost_write_protect                    TRUE      FALSE IMMEDIATE NONE

[oracle@wh ~]$ ./orz.sh _param | grep bmr

_auto_bmr                                enabled

_auto_bmr_req_timeout                    60

_auto_bmr_sess_threshold                 30

_auto_bmr_pub_timeout                    10

_auto_bmr_fc_time                        60

_auto_bmr_bg_time                        3600

_auto_bmr_sys_threshold                  100

_auto_bmr_max_rowno                      1024

_dra_bmr_number_threshold                1000

_dra_bmr_percent_threshold               10

-- 查询 block size

SQL> show parameter db_block_size

NAME                                 TYPE        VALUE

------------------------------------ ----------- ------------------------------

db_block_size                        integer     8192

-- 在pdb中创建测试表

select name from v$pdbs;

alter session set container=POC_PDB;

create table ambr tablespace users as select * from dba_objects;

-- 查询 测试表的 block

select rowid, dbms_rowid.rowid_block_number(rowid) blockno, dbms_rowid.rowid_relative_fno(rowid) fno

from ambr

where rownum < 100

/

ROWID                 BLOCKNO        FNO

------------------ ---------- ----------

AAAWcWAAJAAAACEAAM        132          9

AAAWcWAAJAAAACEAAN        132          9

-- 关闭ambr特性

alter session set Container=CDB$ROOT;

alter system set "_auto_bmr"=disabled;

-- 

dd if=/dev/zero of=/u01/oradata/poc/poc_pdb/poc_pdb_users01.dbf count=1 seek=132 bs=8192 conv=notrunc

-- 

alter session set Container=poc_pdb;

SQL> select count(*) from ambr;

select count(*) from ambr

       *

ERROR at line 1:

ORA-01578: ORACLE data block corrupted (file # 9, block # 132)

ORA-01110: data file 9: '/u01/oradata/poc/poc_pdb/poc_pdb_users01.dbf'

-- 如果没有出现块折断问题,则 alter system flush buffer_cache; 一次.

SQL> drop table ambr purge; -- 由于没有rman备份,不能做块恢复,只能删除问题表.

Table dropped.

-- 开启ambr功能

alter session set Container=CDB$ROOT;

alter system set "_auto_bmr"=enabled;

-- 在pdb中创建测试表

select name from v$pdbs;

alter session set container=POC_PDB;

create table ambr tablespace users as select * from dba_objects;

-- 查询 测试表的 block

select rowid, dbms_rowid.rowid_block_number(rowid) blockno, dbms_rowid.rowid_relative_fno(rowid) fno

from ambr

where rownum < 100

/

ROWID                 BLOCKNO        FNO

------------------ ---------- ----------

AAAWcWAAJAAAACEAAM        132          9

AAAWcWAAJAAAACEAAN        132          9

-- 

dd if=/dev/zero of=/u01/oradata/poc/poc_pdb/poc_pdb_users01.dbf count=1 seek=132 bs=8192 conv=notrunc

--

alter session set container=POC_PDB;

SQL> select count(*) from ambr;

  COUNT(*)

----------

     90991

-- 启用 _auto_bmr 后, 执行 select count(*) from ambr; 后台日志的如下;
- 从日志可以看到, 修复block 是由 ABMR 进程实现的.
- ABMR 只有在出现block块需要修复才会启动.
- ABMR 进程修复完成后,不会立即消失,会保持一段时间.

ALTER SYSTEM SET _auto_bmr='ENABLED' SCOPE=BOTH;

2017-02-20 14:41:43.409000 +08:00

Hex dump of (file 9, block 132) in trace file /u01/app/oracle/diag/rdbms/poc_wh/poc/trace/poc_ora_25255.trc

Corrupt block relative dba: 0x02400084 (file 9, block 132)

Completely zero block found during user buffer read

Reading datafile '/u01/oradata/poc/poc_pdb/poc_pdb_users01.dbf' for corruption at rdba: 0x02400084 (file 9, block 132)

Reread (file 9, block 132) found same corrupt data (no logical check)

Starting background process ABMR

ABMR started with pid=33, OS id=25711 

Automatic block media recovery service is active.

Automatic block media recovery requested for (file# 9, block# 132)

Automatic block media recovery successful for (file# 9, block# 132)

Automatic block media recovery successful for (file# 9, block# 132)

[oracle@wh ~]$ ps -ef | grep 25711

oracle   25711     1  0 14:41 ?        00:00:00 ora_abmr_poc