Beiträge getaggt mit Real-Time Query
Real-Time Query and Automatic Block Media Recovery in 11gR2
Yesterday, I posted about dealing with Block Corruption in general. Special case is Block Corruption in a Data Guard Environment, where we introduced the New Feature Automatic Block Media Recovery (ABMR). Yes, another cool abbrevation from the the Oracle Realm 🙂
To demonstrate this, I have created a Data Guard Configuration as described in my whitepaper 11g Data Guard on the command line, that you may get from the Downloads page. It’s still 11gR1 but can easily be adopted for 11gR2. Will publish an 11gR2 version of it there soon. My scenario looks like this:
SYS@prima > select * from v$version; BANNER -------------------------------------------------------------------------------- Oracle Database 11g Enterprise Edition Release 11.2.0.2.0 - 64bit Production PL/SQL Release 11.2.0.2.0 - Production CORE 11.2.0.2.0 Production TNS for Linux: Version 11.2.0.2.0 - Production NLSRTL Version 11.2.0.2.0 - Production DGMGRL> show configuration; Configuration - myconf Protection Mode: MaxAvailability Databases: prima - Primary database physt - Physical standby database Fast-Start Failover: DISABLED Configuration Status: SUCCESS DGMGRL> show database physt; Database - physt Role: PHYSICAL STANDBY Intended State: APPLY-ON Transport Lag: 0 seconds Apply Lag: 0 seconds Real Time Query: ON Instance(s): physt Database Status: SUCCESS
The first 11g New Feature related to Block Corruption in a Data Guard Environment is less striking than ABMR and similar as restoring blocks from Flashback Logs: If we have a Physical Standby Database present, the corrupted blocks can be restored from there very fast. I corrupted the emp table of scott in the same way as in the previous posting. After using RMAN to validate as shown also already, I do the Blockrecovery
RMAN> blockrecover corruption list;
Starting recover at 17-NOV-10
using channel ORA_DISK_1
finished standby search, restored 1 blocks
starting media recovery
media recovery complete, elapsed time: 00:00:01
Finished recover at 17-NOV-10
I need to emphasize that this restore of blocks from the Physical Standby does not require Real-Time Query there – so we don’t need to purchase Active Data Guard for that. But ABMR does. First ABMR scenario: Block Corruption happens on the Primary Database. I do the same steps as shown in the previous posting to corrupt the block containing the emp table of scott. Now scott connects and selects on the emp table. After a short (about 1 sec) break, the select produces the correct result set! The alert log file of the Primary Database records the following:
Wed Nov 17 08:54:29 2010 Hex dump of (file 4, block 131) in trace file /home/oracle/prima/diag/rdbms/prima/prima/trace/prima_ora_5169.trc Corrupt block relative dba: 0x01000083 (file 4, block 131) Bad header found during buffer read Data in bad block: type: 0 format: 2 rdba: 0xffc00000 last change scn: 0x0000.00000000 seq: 0x0 flg: 0x00 spare1: 0x0 spare2: 0x0 spare3: 0x0 consistency value in tail: 0x00000000 check value in block header: 0xff66 block checksum disabled Reading datafile '/home/oracle/prima/users01.dbf' for corruption at rdba: 0x01000083 (file 4, block 131) Reread (file 4, block 131) found same corrupt data (no logical check) Starting background process ABMR Wed Nov 17 08:54:29 2010 ABMR started with pid=33, OS id=5174 Automatic block media recovery service is active. Automatic block media recovery requested for (file# 4, block# 131) Wed Nov 17 08:54:30 2010 Automatic block media recovery successful for (file# 4, block# 131) Automatic block media recovery successful for (file# 4, block# 131) WARNING: AutoBMR fixed mismatched on-disk single block ffc00000 with in-mem rdba 1000083.
An end user will probably not even notice the Block Corruption on the Primary Database. Second ABMR scenario: Block Corruption happens on the Physical Standby Database. I destroy the block of the emp table in the same way as on the Primary before. Scott now connects on the Physical Standby and does a select on the emp table. Again a short break – no error message but the correct result set is returned! Alert Log of the Physical Standby records:
Wed Nov 17 08:58:10 2010 Hex dump of (file 4, block 131) in trace file /home/oracle/physt/diag/rdbms/physt/physt/trace/physt_ora_5281.trc Corrupt block relative dba: 0x01000083 (file 4, block 131) Bad header found during buffer read Data in bad block: type: 0 format: 2 rdba: 0xffc00000 last change scn: 0x0000.00000000 seq: 0x0 flg: 0x00 spare1: 0x0 spare2: 0x0 spare3: 0x0 consistency value in tail: 0x00000000 check value in block header: 0xff66 block checksum disabled Reading datafile '/home/oracle/physt/users01.dbf' for corruption at rdba: 0x01000083 (file 4, block 131) Reread (file 4, block 131) found same corrupt data (no logical check) Automatic block media recovery requested for (file# 4, block# 131) Wed Nov 17 08:58:11 2010 Automatic block media recovery requested for (file# 4, block# 131) Automatic block media recovery successful for (file# 4, block# 131) Automatic block media recovery successful for (file# 4, block# 131)
Conclusion: We have just seen another benefit of Real-Time Query. Apart from it’s main purpose of making it possible to offload Read-Only (or even Read-Mostly) Applications to the Physical Standby Database, it is now in 11gR2 also leveraged to resolve Block Corruption fast and transparently. We call this 11gR2 New Feature Automatic Block Media Recovery (ABMR)
Real-Time Query for Read-Mostly Applications
Here is one nice little demonstration from my last 11gR2 Data Guard class that I’d like to share with the Oracle Community. It is about enabling Read-Mostly Applications to run on a Physical Standby database, leveraging the 11g New Feature Real-Time Query (part of the Active Data Guard Option). One of the main 11g New Features is the possibility to run Reports anytime on our Physical Standby while it is doing Redo Apply. I have posted about that already here. Now what if we have an Application that is mainly reading but needs to write also to a small amount? That is a problem at first sight:
DGMGRL> show configuration Configuration - myconf Protection Mode: MaxAvailability Databases: prima - Primary database physt - Physical standby database Fast-Start Failover: DISABLED Configuration Status: SUCCESS DGMGRL> show database physt Database - physt Role: PHYSICAL STANDBY Intended State: APPLY-ON Transport Lag: 0 seconds Apply Lag: 0 seconds Real Time Query: ON Instance(s): physt Database Status: SUCCESS
This is my (11gR2) Data Guard Configuration. Now my „Read-Mostly Application“ – enter Scott:
SYS@prima > create user scott identified by tiger; User created. SYS@prima > alter user scott quota unlimited on users; User altered. SYS@prima > grant create session,create table to scott; Grant succeeded. SYS@prima > connect scott/tiger@prima Connected. SCOTT@prima> CREATE TABLE DEPT (DEPTNO NUMBER(2), DNAME VARCHAR2(14), LOC VARCHAR2(13) ); INSERT INTO DEPT VALUES (10, 'ACCOUNTING', 'NEW YORK'); INSERT INTO DEPT VALUES (20, 'RESEARCH', 'DALLAS'); INSERT INTO DEPT VALUES (30, 'SALES', 'CHICAGO'); INSERT INTO DEPT VALUES (40, 'OPERATIONS', 'BOSTON'); commit; Commit complete. SCOTT@prima > create table flag (database_role varchar2(30), start_date date, end_date date); Table created.
My Read-Mostly Applications inserts into the flag-table, then run a „huge“ report and inserts into the flag table again:
SCOTT@prima > @report PL/SQL procedure successfully completed. 1 row created. Commit complete. DEPTNO DNAME LOC ---------- -------------- ------------- 10 ACCOUNTING NEW YORK 20 RESEARCH DALLAS 30 SALES CHICAGO 40 OPERATIONS BOSTON 1 row created. Commit complete.
The content of the report.sql script:
SCOTT@prima > host cat report.sql vari c char(30) exec select SYS_CONTEXT('USERENV','DATABASE_ROLE') into :c from dual; insert into flag values(:c, sysdate,null); commit; select * from dept; insert into flag values(:c, null, sysdate); commit;
This is right now not possible to run on the Physical Standby because of the insert commands:
SCOTT@prima > connect scott/tiger@physt Connected. SCOTT@physt > @report PL/SQL procedure successfully completed. insert into flag values(:c, sysdate,null) * ERROR at line 1: ORA-16000: database open for read-only access Commit complete. DEPTNO DNAME LOC ---------- -------------- ------------- 10 ACCOUNTING NEW YORK 20 RESEARCH DALLAS 30 SALES CHICAGO 40 OPERATIONS BOSTON insert into flag values(:c, null, sysdate) * ERROR at line 1: ORA-16000: database open for read-only access Commit complete.
We need to enable that in an Application-Transparent way. Therefore, we create a Dummy-user to switch to silently when Scott accesses the Physical Standby:
SYS@prima > grant create session to rmostly identified by rmostly;
Grant succeeded.
SYS@prima > grant select on scott.dept to rmostly;
Grant succeeded.
SYS@prima > grant all on scott.flag to rmostly;
Grant succeeded.
SYS@prima > grant create synonym to rmostly;
Grant succeeded.
The user rmostly now gets synonyms with the same name as the Original Applications table. The table that need to get inserts is reached via a database link that points to the Primary – modifying it there.
SYS@prima > create public database link prod connect to scott identified by tiger using 'prod'; Database link created. SYS@prima > connect rmostly/rmostly@prima Connected. RMOSTLY@prima > create synonym dept for scott.dept; Synonym created. RMOSTLY@prima > create synonym flag for scott.flag@prod; Synonym created.
The used connect descriptor prod should always lead to the Primary Database even after Role Changes. We will discuss that later on in more detail. Now we need a Logon Trigger that silently switches to the rmostly schema if Scott connects to the Physical Standby:
SYS@prima > CREATE or replace TRIGGER switch_schema_trigger
AFTER LOGON ON scott.schema
BEGIN
IF (SYS_CONTEXT('USERENV','DATABASE_ROLE')
IN ('PHYSICAL STANDBY'))
THEN
execute immediate
'alter session set current_schema = rmostly';
END IF;
end;
/
Trigger created.
Also, we need a connect descriptor that always connects Scott to the Physical Standby regardless of Role Changes. Now we are going to setup this part:
SYS@prima > exec dbms_service.create_service('prod','prod')
PL/SQL procedure successfully completed.
SYS@prima > exec dbms_service.create_service('rmostly','rmostly')
PL/SQL procedure successfully completed.
SYS@prima > exec dbms_service.start_service('prod')
PL/SQL procedure successfully completed.
SYS@prima > @physt
Connected.
SYS@physt > exec dbms_service.start_service('rmostly')
PL/SQL procedure successfully completed.
Please notice that the rmostly service was started on the Physical Standby. For now, the services are on the appropriate Database, but the following trigger ensures that this will be still the case after Role Changes like switchover:
SYS@prima > CREATE or replace TRIGGER service_management AFTER STARTUP ON DATABASE DECLARE VROLE VARCHAR(30); VOPEN_MODE VARCHAR(30); BEGIN SELECT DATABASE_ROLE INTO VROLE FROM V$DATABASE; SELECT OPEN_MODE INTO VOPEN_MODE FROM V$DATABASE; IF VROLE = 'PRIMARY' THEN begin DBMS_SERVICE.START_SERVICE ('PROD'); DBMS_SERVICE.STOP_SERVICE ('RMOSTLY'); end; ELSIF VROLE = 'PHYSICAL STANDBY' THEN begin IF VOPEN_MODE like 'READ ONLY%' THEN DBMS_SERVICE.START_SERVICE ('RMOSTLY'); END IF; DBMS_SERVICE.STOP_SERVICE ('PROD'); end; END IF; END; / Trigger created.
The tnsnames.ora used by the Application Clients looks like this:
$ cat $ORACLE_HOME/network/admin/tnsnames.ora
PROD =
(DESCRIPTION =
(ADDRESS_LIST =
(ADDRESS = (PROTOCOL = TCP)(HOST = uhesse1)(PORT = 1521))
(ADDRESS = (PROTOCOL = TCP)(HOST = uhesse2)(PORT = 1521))
)
(CONNECT_DATA =
(SERVER = DEDICATED)
(SERVICE_NAME = prod)
)
)
RMOSTLY =
(DESCRIPTION =
(ADDRESS_LIST =
(ADDRESS = (PROTOCOL = TCP)(HOST = uhesse1)(PORT = 1521))
(ADDRESS = (PROTOCOL = TCP)(HOST = uhesse2)(PORT = 1521))
)
(CONNECT_DATA =
(SERVER = DEDICATED)
(SERVICE_NAME = rmostly)
)
)
In effect, an Application Client connecting to the Physical Standby with RMOSTLY can insert into the Primary table while selecting from the Physical Standby. We assume, that the main part of that Application is doing select, so we have still offloaded the major work from the Primary to the Physical Standby:
SCOTT@rmostly > alter session set nls_date_format='hh24:mi:ss'; Session altered. SCOTT@rmostly > @report PL/SQL procedure successfully completed. 1 row created. Commit complete. DEPTNO DNAME LOC ---------- -------------- ------------- 10 ACCOUNTING NEW YORK 20 RESEARCH DALLAS 30 SALES CHICAGO 40 OPERATIONS BOSTON 1 row created. Commit complete. SCOTT@rmostly > select * from flag; DATABASE_ROLE START_DA END_DATE ------------------------------ -------- -------- PRIMARY 14:24:35 PRIMARY 14:24:35 PHYSICAL STANDBY 14:53:21 PHYSICAL STANDBY 14:53:21
The SQL prompt BTW changes accordingly because of this little login.sql:
SCOTT@rmostly > host cat login.sql set lines 800 set pages 300 column name format a50 column file_name format a50 column client_info format a10 column machine format a20 column MASTER_LINK format a10 column parameters format a10 column def_parameters format a10 SET SQLPROMPT "_USER'@'_CONNECT_IDENTIFIER > "
Conclusion: We can even offload Read-Mostly Applications to Physical Standby Databases in 11g with little effort!
Real-Time Query in 11gR2
You may have noticed the spectacular 11g (R1) New Feature Real-Time Query already, that makes it possible to use a Physical Standby Database for Reporting while it is still applying Redo-Protocol received from the Primary Database. There are some additions to that feature in 11gR2.
DGMGRL> show configuration
Configuration - myconf
Protection Mode: MaxAvailability
Databases:
prima - Primary database
physt - Physical standby database
Fast-Start Failover: DISABLED
Configuration Status:
SUCCESS
DGMGRL> show database physt
Database - physt
Role: PHYSICAL STANDBY
Intended State: APPLY-ON
Transport Lag: 0 seconds
Apply Lag: 0 seconds
Real Time Query: ON
Instance(s):
physt
Database Status:
SUCCESS
I think the red sections above are new in 11gR2 and make it easier to recognize that we are using Real-Time Query – therefore we may want to license Active Data Guard, if not done already 🙂
Now let’s suppose we make use of the Physical Standby for Reporting but want to make sure that the Reports are up to date with the Primary Changes:
SQL> select * from v$version; BANNER -------------------------------------------------------------------------------- Oracle Database 11g Enterprise Edition Release 11.2.0.1.0 - Production PL/SQL Release 11.2.0.1.0 - Production CORE 11.2.0.1.0 Production TNS for Linux: Version 11.2.0.1.0 - Production NLSRTL Version 11.2.0.1.0 - Production SQL> select open_mode from v$database; OPEN_MODE -------------------- READ ONLY WITH APPLY
SQL> col name for a30
col value for a20
select name,value from v$dataguard_stats;
NAME VALUE
------------------------------ --------------------
transport lag +00 00:00:00
apply lag +00 00:00:00
apply finish time +00 00:00:00.000
estimated startup time 25
There is no Apply Lag right now between Primary and Standby. The shown value of OPEN_MODE is also a New Feature in 11gR2, as well as the following parameter:
SQL> connect scott/tiger@physt
Connected.
SQL> alter session set STANDBY_MAX_DATA_DELAY=1;
Session altered.
That setting means that in Scott’s session on the Physical Standby, we will only tolerate a Lag Time of 1 second between the Data on the Standby and the Data on the Primary. A value of zero would require full synchronicity. I will make sure that this condition cannot be met by interrupting the Redo Transport from Primary to Standby from another session:
SQL> connect sys/oracle@prima as sysdba
Connected.
SQL> alter system set log_archive_dest_state_2=defer;
System altered.
Have you noticed that there are now 31 Log Archive Destinations, by the way? I will now change data on the Primary while the Standby cannot apply, thereby creating a Transport Lag and also an Apply Lag. It’s easier for me to demonstrate it that way as by generating a so high load on the Primary that this leads to a significant Apply Lag on the Standby because MRP can’t keep the pace. Same session:
SQL> connect scott/tiger@prima
Connected.
SQL> select * from dept;
DEPTNO DNAME LOC
---------- -------------- -------------
10 ACCOUNTING NEW YORK
20 RESEARCH DALLAS
30 ACCOUNTING CHICAGO
40 OPERATIONS BOSTON
SQL> update dept set dname='A' where deptno=10;
1 row updated.
SQL> commit;
Commit complete.
Back to the Scott session on the Physical Standby:
SQL> select * from dept;
select * from dept
*
ERROR at line 1:
ORA-00604: error occurred at recursive SQL level 1
ORA-03172: STANDBY_MAX_DATA_DELAY of 1 seconds exceeded
The Standby is lagging too far behind the demanded maximum Lag Time. In 11gR1, this would have given the old result set like that:
SQL> alter session set STANDBY_MAX_DATA_DELAY=none; Session altered. SQL> select * from dept; DEPTNO DNAME LOC ---------- -------------- ------------- 10 ACCOUNTING NEW YORK 20 RESEARCH DALLAS 30 ACCOUNTING CHICAGO 40 OPERATIONS BOSTON
Another 11gR2 New Feature is the option to synchronize the Standby explicitly before a query:
SQL> alter session set STANDBY_MAX_DATA_DELAY=1; Session altered.
SQL> alter session sync with primary;
ERROR:
ORA-03173: Standby may not be synced with primary
SQL> host oerr ora 3173
03173, 00000, "Standby may not be synced with primary"
// *Cause: ALTER SESSION SYNC WITH PRIMARY did not work because either the
// standby was not synchronous or in a data-loss mode with regard
// to the primary, or the standby recovery was terminated.
// *Action: Make the standby synchronous and no-data-loss with regard
// to the primary. Make the standby recovery active.
The SYNC is not possible because of the DEFER I did above. I renable Redo Transport and then try again:
SQL> connect sys/oracle@prima as sysdba
Connected.
SQL> alter system set log_archive_dest_state_2=enable;
System altered.
SQL> connect scott/tiger@physt
Connected.
SQL> alter session set STANDBY_MAX_DATA_DELAY=1;
Session altered.
SQL> alter session sync with primary;
Session altered.
Conclusion: With the 11gR2 addition STANDBY_MAX_DATA_DELAY, we can now enforce a certain level of synchronicity during the usage of Real-Time Query in order to make the results we get from the Physical Standby reliable.
A Practical Guide to Data Guard: Real-Time Query https://t.co/77dmMULkZF #Oracle Learning Streams: FREE for OCP, OCE pic.twitter.com/5Nv9v3Q9OX
— Uwe Hesse (@UweHesse) October 23, 2015