Rollback segment Damaged Repair instance

A few days ago, my colleague called to say that the company's development bank died 4 or 5 times a day.
I just got back to the company today.

Check Alertlog and find hundreds of errors, all of which are:
Errors in file D:\oracle\admin\croot\bdump\crootSMON.TRC:
Ora-01578:oracle data Block corrupted (file # 14715)
Ora-01110:data file: ' D:\ORACLE\ORADATA\CROOT\TRBS1. ORA '

The file#=22 file is a data file in the rollback segment table space, and it is obvious that the rollback segment data block is corrupted.
However, the database can be used, it can be opened normally, but not normal shutdown,
When closing the newspaper:
Shutting down instance (immediate)
License High Water mark = 90
Mon Oct 25 12:03:23 2004
ORACLE Instance croot (PID = 6)-Error 1578 encountered while recovering transaction (2, 50).
Mon Oct 25 12:03:23 2004
Errors in file D:\oracle\admin\croot\bdump\crootSMON.TRC:
Ora-01578:oracle data Block corrupted (file # 14715)
Ora-01110:data file: ' D:\ORACLE\ORADATA\CROOT\TRBS1. ORA '

And the whole case is crash.

Check the bad block where the object, found to be RBS01, and then check the Dba_rollback_segs view, found that in addition to RBS01 is online state, the other is partitial available state, manual will be the rest of the rollback segment online after Can not be offline again, immediately have active transaction occupy rollback segment, all other sessions stopped, this is still the case, look at the v$trasaction view, no records show that there are running transactions.

This situation, can not normally drop off the rollback section, naturally there is no way to drop off the rollback segment table space.

Since it is a development library, it is in a non-archive mode and there is no scheduled backup, so it is not necessary to restore the database.

But also good is the development storehouse, those affairs lost also does not matter at all.
So, close the database, modify the initialization parameter file, and add suppressed parameters:
_corrupted_rollback_segments= (RBS1,RBS2,RBS4,RBS5,RBS6,RBS7,RBS8,RBS9,RBS10)

All flags are corrupted, then start the database to Mount state
Drop off the file#=22 data file
Then recover database;alter database open;

After the database is opened, check the V$rollname view again and find that only the rollback segment in the system table space exists, OK, then go to the Dba_rollback_segs view check state, and find that other than the system rollback segment is in the need recover state, OK That's right. Start the deletion.

Drop all the rollback segments directly, drop the rollback segment table space, create a new table space, create a new rollback segment, and make all rollback segments online, with a minor episode in the middle, which is mentioned later.

Finally, the database is normally closed, the init parameters are modified, the _corrupted_rollback_segments line is commented out, and the database is finally started normally. Look for a few tables, make a few insert test, no problem, OK, repair completed.

Episode: A problem occurred when creating a new rollback segment, delayed for some time, because it is 8i library, the system table space is DMT, so the rollback segment table space must also be DMT, began to create a LMT table space according to custom, and then create a rollback segment is always an error, said the non-system rollback segment can not use System tablespace, think for a while, just think of the DMT and LMT this matter, and then drop off the new LMT table space, create a DMT table space, and then recreate the rollback segment, no problem.