11.2 Data Guard Physical Standby Switchover Best Practices using SQL*Plus [ID 1304939.1]

11.2 Data Guard Physical Standby Switchover Best Practices using SQL*Plus [ID 1304939.1]
		Modified 21-MAR-2011     Type HOWTO     Status PUBLISHED

In this Document
  Goal
  Solution
     I. Prerequisites / Preparation
     II. Pre-Switchover Checks
     III. Switchover
     IV. Post-Switchover Steps

---

 

 

Applies to:

Oracle Server - Enterprise Edition - Version: 11.2.0.1 and later   [Release: 11.2 and later ]
Information in this document applies to any platform.

Goal

Perform trouble free Data Guard switchover.

Note: for Data Guard switchover using the Broker please refer to Note 1305019.1 - "11.2 Data Guard Physical Standby Switchover Best Practices using the Broker"

Solution

I. Prerequisites / Preparation

These are items that should only have to be done once during configuration and setup.

Alert: If you upgraded your databases to 11.2.0.2 from a prior release (e.g. 10.2, 11.1, 11.2.0.1) it is imperative that you refer to Note 1288640.1 "Managed Recovery (MRP) Fails w/ ORA-328 After Upgrade to 11.2.0.2 and Switchover" before continuing.

Apply Latest Patch Bundle.

• Review Note 785351.1 "11gR2 Upgrade Companion"
  • Make sure to check the “Patches Recommended” tab.
• See Note 756671.1 for the latest available patches or patchset updates.

Review Primary Database Initialization Parameters

Ensure that the LOG_ARCHIVE_CONFIG & DG_CONFIG initialization parameters are established at the primary database. See Section 3.1.4 Set Primary Database Initialization Parameters

 

Verify the Setup

Follow the steps at Section 3.2.7 Verify the Physical Standby Database Is Performing Properly.

 

Understand and Test Fallback Options

See Appendix A.4 Problems Switching Over to a Physical Standby Database

II. Pre-Switchover Checks

These steps should be completed before the switchover planned maintenance window begins. Our recommendation is that these are done a couple days in advance.

Verify Managed Recovery is running on the standby

The following query at the standby verifies that managed recovery is running:

SQL> SELECT PROCESS FROM V$MANAGED_STANDBY WHERE PROCESS LIKE 'MRP%';

The following query at the Primary verifies that recovery is running with “REAL TIME APPLY” option. In the example below, LOG_ARCHIVE_DEST_2 is established to ship redo to the target standby (dest_id=2):

SQL> SELECT RECOVERY_MODE FROM V$ARCHIVE_DEST_STATUS WHERE DEST_ID=2;

RECOVERY_MODE
-----------------------
MANAGED REAL TIME APPLY

If managed standby recovery is not running or not started with real-time apply, restart managed recovery with real-time apply enabled:

SQL> ALTER DATABASE RECOVER MANAGED STANDBY DATABASE CANCEL;
SQL> ALTER DATABASE RECOVER MANAGED STANDBY DATABASE USING CURRENT LOGFILE DISCONNECT;

Note: If you previously defined a delay for this standby the delay is ignored when you start real time apply
For more information see Section 3.2.7 Verify the Physical Standby Database Is Performing Properly

Ensure Online Redo Log Files on the Target Physical Standby have been cleared

Online redo logs on the target physical standby need to be cleared before that standby database can become a primary database. Although this will automatically happen as part of the SWITCHOVER TO PRIMARY command, it is recommended that the logs are cleared prior to the switchover.

Setting the LOG_FILE_NAME_CONVERT parameter at the physical standby will cause the online redo logs to be automatically cleared when managed recovery is started on the standby.

If your databases are using Oracle Managed Files (OMF) or you have already set the parameter LOG_FILE_NAME_CONVERT you can skip this step as the online log files will always be cleared automatically.

Clearing online redo logs as part of the SWITCHOVER TO PRIMARY command can make the switchover command susceptible to termination by another process that is waiting on access to the CONTROLFILE. The CONTROLFILE waiter will attempt to kill the switchover after a timeout is 15 minutes.

Oracle recommends setting LOG_FILE_NAME_CONVERT to automatically clear online redo logs on the physical standby database. In the event the primary database and the physical standby database have the exact same directory path to the online redo logs, it is acceptable to set LOG_FILE_NAME_CONVERT such that the entry pairs have the same value.

As an example, if the online redo logs are stored in /oradata/order_db/redo for both the primary and physical standby databases on their respective servers, you can set the parameter value as

LOG_FILE_NAME_CONVERT=’/oradata/order_db/redo/’,’/oradata/order_db/redo/’

This will initiate automatic clearing of the online redo logs on the physical standby database when managed recovery is started.

Since the LOG_FILE_NAME_CONVERT parameter is not dynamic you must restart the standby database for the property change to take affect.

If you have not set your environment to automatically clear the online redo logs and you do not want to restart the standby database, you should manually clear them at some point prior to the switchover. This can be done at any time.

On the target physical standby run the following query to determine if the online redo logs have not been cleared:

SQL> SELECT DISTINCT L.GROUP# FROM V$LOG L, V$LOGFILE LF
          WHERE L.GROUP# = LF.GROUP#
          AND L.STATUS NOT IN (‘UNUSED’, ‘CLEARING’,’CLEARING_CURRENT’);

If the above query returns rows, on the target physical standby issue the following statement for each GROUP# returned:

SQL> ALTER DATABASE CLEAR LOGFILE GROUP <ORL GROUP# from the query above>;

Note that later when you do the actual switchover if it is terminated by a CONTROLFILE waiter timeout, just re-issue the SWITCHOVER TO PRIMARY command until it completes successfully.

You should monitor your alert log to ensure your online redo logs are being cleared and you are not experiencing some other issue.

Verify there are no large Gaps

Identify the current sequence number for each thread on the primary database

SQL> SELECT THREAD#, SEQUENCE# FROM V$THREAD;

Verify the target physical standby database has applied up to, but not including the logs from the primary query. On the standby the following query should be within 1 or 2 of the primary query result.

SQL> SELECT THREAD#, MAX(SEQUENCE#) FROM V$ARCHIVED_LOG
          WHERE APPLIED = 'YES'
          AND RESETLOGS_CHANGE# = (SELECT RESETLOGS_CHANGE#
          FROM V$DATABASE_INCARNATION WHERE STATUS = ‘CURRENT’)
          GROUP BY THREAD#;

If large gaps exist (more than 3 logs) then see Section 6.4.3 Redo Gap Detection and Resolution.

Verify Primary and Standby tempfiles match and all datafiles are ONLINE

For each temporary tablespace on the standby, verify that temporary files associated with that tablespace on the primary database also exist on the standby database. Tempfiles added after initial standby creation are not propagated to the standby. Run this query on both the primary and target physical standby databases and verify that they match.

SQL> SELECT TMP.NAME FILENAME, BYTES, TS.NAME TABLESPACE
           FROM V$TEMPFILE TMP, V$TABLESPACE TS WHERE TMP.TS#=TS.TS#;

If the queries do not match then you can correct the mismatch now or immediately after the open of the new primary database.

Prior to switchover, on the target standby, verify that all datafiles necessary for updates after role transition to primary are ONLINE.

On the target standby:

SQL> SELECT NAME FROM V$DATAFILE WHERE STATUS=’OFFLINE’;

If there are any OFFLINE datafiles, and these are needed after switchover, bring them ONLINE:

SQL> ALTER DATABASE DATAFILE ‘datafile-name’ ONLINE;

 

III. Switchover

These steps are completed as part of the switchover process on the day of the planned outage.

Clear Potential Blocking Parameters & Jobs

Capture current job state on the primary:

SQL> SELECT * FROM DBA_JOBS_RUNNING;

Depending on what the running job is, be ready to terminate the job if necessary.

SQL> SELECT OWNER, JOB_NAME, START_DATE, END_DATE, ENABLED FROM 

           DBA_SCHEDULER_JOBS WHERE ENABLED=’TRUE’ AND OWNER <> ‘SYS”;

SQL> SHOW PARAMETER job_queue_processes

Note: Job candidates to be disabled among others: oracle text sync and optimizer, RMAN backups, application garbage collectors, application background agents.

Block further job submission

SQL> ALTER SYSTEM SET job_queue_processes=0 SCOPE=BOTH SID=’*’;

Disable any jobs that may interfere.

SQL> EXECUTE DBMS_SCHEDULER.DISABLE( <job_name> );

Shutdown all mid-tiers (optional)

This can be done in parallel to the switchover.

$ opmnctl stopall

Note: If using a local standby with an application that is following the MAA “ Client Failover Best Practices for Data Guard 11g Release 2” paper recommendations that ensure the application database service is only active on the primary, this step should be skipped.

 

Turn on Data Guard tracing on primary and standby

Tracing is turned on to have diagnostic information available in case any issues arise. Turning on tracing does not have any noticeable impact on switchover time but does require space for the trace output.

Capture the current value on both the primary and the target physical standby databases

SQL> SHOW PARAMETER log_archive_trace

Set Data Guard trace level to 8191 on both the primary and the target physical standby databases

SQL> ALTER SYSTEM SET log_archive_trace=8191;

Trace output will appear under the destination pointed to by the database parameter BACKGROUND_DUMP_DEST with “mrp” in the file name.

Tail Alert Logs (optional) on all instances

Locate alert logs by showing database parameter background_dump_dest

SQL> SHOW PARAMETER background_dump_dest

Tail the alert logs

> tail –f <background_dump_dest location>/alert*

Create Guaranteed Restore Points (optional)

The standard switchover fallback options should suffice for successfully backing out of a switchover. However, if you want an additional fallback option then you can create a guaranteed restore point on the primary and standby database participating in the switchover.

On the standby

Stop the apply process

SQL> ALTER DATABASE RECOVER MANAGED STANDBY DATABASE CANCEL;

Create a guaranteed restore point

SQL> CREATE RESTORE POINT SWITCHOVER_START_GRP GUARANTEE FLASHBACK DATABASE;

Start the apply process

SQL> ALTER DATABASE RECOVER MANAGED STANDBY DATABASE
USING CURRENT LOGFILE DISCONNECT;

On the primary

Create a guaranteed restore point

SQL> CREATE RESTORE POINT SWITCHOVER_START_GRP GUARANTEE FLASHBACK DATABASE;

Note: If a guaranteed restore points are created, make sure they are dropped post-switchover!

Verify that the primary database can be switched to the standby role

Query the SWITCHOVER_STATUS column of the V$DATABASE view on the primary database:

SQL> SELECT SWITCHOVER_STATUS FROM V$DATABASE;

SWITCHOVER_STATUS
-----------------
TO STANDBY

A value of TO STANDBY or SESSIONS ACTIVE (which requires the WITH SESSION SHUTDOWN clause on the switchover command) indicates that the primary database can be switched to the standby role. If neither of these values is returned, a switchover is not possible because redo transport is either mis-configured or is not functioning properly. See Appendix A.4 Problems Switching Over to a Physical Standby Database

If The Primary is a RAC, then shutdown all secondary primary instances

A normal or immediate shutdown can be done, but to expedite the shutdown issue a SHUTDOWN ABORT on secondary RAC instances on the primary cluster only leaving one Primary instance up. Wait until the remaining Primary instance has completed cluster reconfiguration (and performed recovery if you chose to abort the secondary instances) before continuing.

Switchover the primary to a standby database

SQL> ALTER DATABASE COMMIT TO SWITCHOVER TO STANDBY WITH SESSION SHUTDOWN;

If an ORA-16139 error is encountered, as long as V$DATABASE.DATABASE_ROLE=’PHYSICAL STANDBY’, then you can proceed. A common case where this can occur is when there are a large number of data files. Once managed recovery is started on the new standby, the database will recover.

If the role was not changed then you need to cancel the switchover and review the alert logs and trace files further.

Verify the standby has received the end-of-redo (EOR) log(s)

In the primary alert log you will see messages like these:

Switchover: Primary controlfile converted to standby controlfile succesfully.
Tue Mar 15 16:12:15 2011
MRP0 started with pid=17, OS id=2717
MRP0: Background Managed Standby Recovery process started (SFO)
Serial Media Recovery started
Managed Standby Recovery not using Real Time Apply
Online logfile pre-clearing operation disabled by switchover
Media Recovery Log /u01/app/flash_recovery_area/SFO/archivelog/2011_03_15/o1_mf_1_133_6qzl0yvd_.arc
Identified End-Of-Redo for thread 1 sequence 133
Resetting standby activation ID 0 (0x0)
Media Recovery End-Of-Redo indicator encountered
Media Recovery Applied until change 4314801
MRP0: Media Recovery Complete: End-Of-REDO (SFO)
MRP0: Background Media Recovery process shutdown (SFO)
Tue Mar 15 16:12:21 2011
Switchover: Complete - Database shutdown required (SFO)
Completed: ALTER DATABASE COMMIT TO SWITCHOVER TO PHYSICAL STANDBY WITH SESSION SHUTDOWN

And correspondingly in the standby alert log file you should see messages like these:

Tue Mar 15 16:12:15 2011
RFS[8]: Assigned to RFS process 2715
RFS[8]: Identified database type as 'physical standby': Client is Foreground pid 2568
Media Recovery Log /u01/app/flash_recovery_area/NYC/archivelog/2011_03_15/o1_mf_1_133_6qzl0yjp_.arc
Identified End-Of-Redo for thread 1 sequence 133
Resetting standby activation ID 2680651518 (0x9fc77efe)
Media Recovery End-Of-Redo indicator encountered
Media Recovery Continuing
Resetting standby activation ID 2680651518 (0x9fc77efe)
Media Recovery Waiting for thread 1 sequence 134

In versions prior to Oracle Database 11g Release 2, the MRP (Redo Apply coordinator) would stop automatically after processing the End-of-Redo marker. With Oracle Database 11g Release 2, it no longer stops leaving all bystander standby databases still ready to apply redo from the new primary database without having to be restarted. The MRP process will be shut down automatically by the switchover command when executed at the target standby database.

Verify that the standby database can be switched to the primary role

Query the SWITCHOVER_STATUS column of the V$DATABASE view on the standby database:

SQL> SELECT SWITCHOVER_STATUS FROM V$DATABASE;

SWITCHOVER_STATUS
-----------------
TO PRIMARY

A value of TO PRIMARY or SESSIONS ACTIVE indicates that the standby database is ready to be switched to the primary role. If neither of these values is returned, verify that redo apply is active and that redo transport is configured and working properly. Continue to query this column until the value returned is either TO PRIMARY or SESSIONS ACTIVE.

Switchover the standby database to a primary

SQL> ALTER DATABASE COMMIT TO SWITCHOVER TO PRIMARY WITH SESSION SHUTDOWN;

In the standby alert log file you should see messages like these:

Tue Mar 15 16:16:44 2011
ALTER DATABASE COMMIT TO SWITCHOVER TO PRIMARY WITH SESSION SHUTDOWN
ALTER DATABASE SWITCHOVER TO PRIMARY (NYC)
Maximum wait for role transition is 15 minutes.
Switchover: Media recovery is still active
Role Change: Canceling MRP - no more redo to apply
Tue Mar 15 16:16:45 2011
MRP0: Background Media Recovery cancelled with status 16037
Errors in file /u01/app/diag/rdbms/nyc/NYC/trace/NYC_pr00_2467.trc:
ORA-16037: user requested cancel of managed recovery operation
Managed Standby Recovery not using Real Time Apply
Recovery interrupted!
Waiting for MRP0 pid 2460 to terminate
Errors in file /u01/app/diag/rdbms/nyc/NYC/trace/NYC_pr00_2467.trc:
ORA-16037: user requested cancel of managed recovery operation
Tue Mar 15 16:16:45 2011
MRP0: Background Media Recovery process shutdown (NYC)
Role Change: Canceled MRP

Open the new primary database

SQL> ALTER DATABASE OPEN;

Note: There will be an increase in I/O activity while the new primary’s standby redo logs are cleared.

Correct any tempfile mismatch

If there was a tempfile that was not corrected during the pre-switchover check, then correct it now on the new primary.

Restart the new standby

If the new standby database (former primary database) was not shutdown since switching it to standby, bring it to the mount state and start managed recovery. This can be done in parallel to the new primary open.

SQL> SHUTDOWN ABORT;

Note: If you use IMMEDIATE, an ABORT will be performed anyway as of 11.2.0.2 and you would see the following in the alert log:

Performing implicit shutdown abort due to switchover to physical standby
Shutting down instance (abort)
License high water mark = 15
USER (ospid: 14665): terminating the instance
Instance terminated by USER, pid = 14665

SQL> STARTUP MOUNT;

SQL> ALTER DATABASE RECOVER MANAGED STANDBY DATABASE USING CURRENT LOGFILE DISCONNECT;

Note: If you were using a delay for your standby then you would restart the apply without real time apply:

SQL> ALTER DATABASE RECOVER MANAGED STANDBY DATABASE DISCONNECT;

Finally, if the database is a RAC, then start all secondary instances on the new standby.

Contingency or Fallback

See Appendix A.4.5 Roll Back After Unsuccessful Switchover and Start Over in the Data Guard Concepts and Administration manual.

IV. Post-Switchover Steps

Set Trace to Prior Value

For each instance on the Primary and Standby:

SQL> ALTER SYSTEM SET log_archive_trace=<prior value>;

Reset Jobs

Set the job queue processes to its original value on the new standby.

SQL> ALTER SYSTEM SET job_queue_processes=<value saved> scope=both sid=’*’

Enable any jobs that were disabled.

SQL> EXECUTE DBMS_SCHEDULER.ENABLE(<for each job name captured>);

Drop any Switchover Guaranteed Restore Points

On all databases where a Guaranteed Restore point was created

SQL> DROP RESTORE POINT SWITCHOVER_START_GRP;