Overview

• This document describes a method to bypass and start the DB first when the Undo Tablespace Datafile is lost in a No Archive Mode environment without backups.
• This method is a forced recovery approach (assuming no valid backups exist), which may result in some data loss depending on the case.
• This operation is a forced recovery task and should be performed after blocking external business connections during the process. (It is recommended to temporarily change the Tibero listener port before proceeding, if possible.)

Method

Undo Loss Recovery Scenario

Deletion of Undo Tablespace Datafile

Check and Delete Datafile

Delete the Undo Tablespace Datafile named undo001.dtf.

Confirm DB Restart

Confirm that the DB starts only up to Mount mode, indicating that the undo datafile (undo001.dtf) requires recovery.

Apply Relevant Parameters

The parameters are as follows:

• _ENABLE_UNDO_TS_OFFLINE=Y
• _SKIP_TX_RECOVERY=Y
• _CF_DD_CHECK=N
   

Perform Recovery Operation

After applying the parameters, start the DB in Mount mode.

Once in Mount mode, set the Undo Tablespace to Offline, then perform the recover operation.

(tbboot mount startup - Declare Undo Tablespace Offline - Perform automatic recovery)

Create New Undo Tablespace

This process creates a new Undo Tablespace and removes the existing one.

Start DB in Normal Mode

You can confirm normal mode startup after temporary parameter addition and automatic recovery.

Change Undo Usgmt

Since the current Undo Tablespace is Offline, it is difficult to check related views for Usgmt change status.

Query the internal table (_DD_TS) and forcibly change the status of the previously offline Undo Tablespace.

Change the status to 3 for the TS_ID of the existing Undo Tablespace named UNDO, so that this Undo is no longer referenced.

Replace Undo Tablespace

Create a new Undo Tablespace (example named UNDO_NEW).

Switch the DB's Undo Tablespace to the new one and remove the existing Undo Tablespace named UNDO.

Remove Temporary Parameters and Restart DB

After forced recovery, remove (or comment out) the parameters applied in step 2-2, shut down the DB, and restart it.

As explained in the overview, since TX Recovery (transaction recovery) has not been performed, some consistency issues may occur.

Remove the parameters added in step 2-2 after DB shutdown.

After startup, confirm that the Undo Tablespace has changed to UNDO_NEW.

Also confirm that there are no additional files requiring recovery.