Undo tablespace corruption will terminate DB operation

Hi All,
Just for the concept, if undotablespace datafile got corrupted due to any reasion, will it force to terminate the DB operation.
O know this will happen with control file and SYTEM datafile only.
Please confirm.
Regards,
SS

Hmm what to check? It says the same, it is not mandatory that instance would surely get shut down with the loss of undo tablespace. But you MUST do the recovery of it when the database is closed. I guess that's what that doc is also saying.
Update:
See below where I would show the same thing happening. We would start from the undo tablespace getting deleted and would check the alert log for any message. Also we would check the status of teh database after the drop,
Session 1
SQL*Plus: Release 10.2.0.1.0 - Production on Thu Aug 6 10:43:35 2009
Copyright (c) 1982, 2005, Oracle. All rights reserved.
Connected to an idle instance.
SQL> startup
ORACLE instance started.
Total System Global Area 167772160 bytes
Fixed Size                  1218316 bytes
Variable Size              79694068 bytes
Database Buffers           83886080 bytes
Redo Buffers                2973696 bytes
Database mounted.
Database opened.
SQL> shut immediate
Database closed.
Database dismounted.
ORACLE instance shut down.
SQL> startup mount
ORACLE instance started.
Total System Global Area 167772160 bytes
Fixed Size                  1218316 bytes
Variable Size              79694068 bytes
Database Buffers           83886080 bytes
Redo Buffers                2973696 bytes
Database mounted.
SQL> alter database archivelog;
Database altered.
SQL> alter database open;
Database altered.
SQL> select name from V$datafile;
NAME
--------------------------------------------------------------------------------/u01/app/oracle/oradata/orcl/system01.dbf
/u01/app/oracle/oradata/orcl/undotbs01.dbf
/u01/app/oracle/oradata/orcl/sysaux01.dbf
/u01/app/oracle/oradata/orcl/users01.dbf
SQL> !rm /u01/app/oracle/oradata/orcl/undotbs01.dbf
SQL> select status from V$instance;
STATUS
------------ <-- The database is open
OPEN
#At this time, in the alert log there was no message.
#Let's try some ddl/dmls.
SQL> create table t( a number);
Table created.
SQL> insert into t values(1);
1 row created.
SQL> commit;
Commit complete.
SQL> delete t;
1 row deleted.
#No problem so far. Let's issue a checkpoint, this would force oracle to check the file and its status.
SQL> alter system checkpoint;
System altered.
SQL>
Session-2(Alert log's tail)
Output snipped
ORA-01110: data file 2: '/u01/app/oracle/oradata/orcl/undotbs01.dbf'
Thu Aug 6 11:11:19 2009
Errors in file /u01/app/oracle/admin/orcl/bdump/orcl_pmon_4402.trc:
ORA-00376: file 2 cannot be read at this time
ORA-01110: data file 2: '/u01/app/oracle/oradata/orcl/undotbs01.dbf'
Thu Aug 6 11:11:19 2009
Errors in file /u01/app/oracle/admin/orcl/bdump/orcl_pmon_4402.trc:
ORA-00376: file 2 cannot be read at this time
ORA-01110: data file 2: '/u01/app/oracle/oradata/orcl/undotbs01.dbf'
Thu Aug 6 11:11:19 2009
Errors in file /u01/app/oracle/admin/orcl/bdump/orcl_pmon_4402.trc:
ORA-00376: file 2 cannot be read at this time
ORA-01110: data file 2: '/u01/app/oracle/oradata/orcl/undotbs01.dbf'
#Let's check the status again,
SQL> select stauts from V$isntance;
select stauts from V$isntance
ERROR at line 1:
ORA-00603: ORACLE server session terminated by fatal error
Its crashed now as db couldn't find the undo tablespace.
HTH
Aman....
Edited by: Aman.... on Aug 6, 2009 11:28 AM added demo.

Similar Messages

UNDO tablespace corrupted

Hi!
I have been having problems for the past couple days with restarting my database. For some reason when I shutdown, once I start it up again the undo tablespaces is offline and needs to be recovered. When the application tries to connect to the database it gives a "ora-01552 : cannot use system rollback segment for non-system tablespace." I do a recover on that tablespace and bring it back up, and it works fine... until the next restart.. Would anyone have an idea on what's the problem, and what I can do to fix it once and for all?
Thanks
Yazil Santoyo

did you try metalink, there are lot of options / suggestions for your particular error.
Is your database in Auto undo mode or manual undo?
Thanks
Gopal

Dropping Undo Tablespace 9i

Hey Folks,
I really need some help here.
I'm trying to drop/recreate the undo tablespace (currently 25gig) Oracle 9i
When I try to drop the undo tablespace 'UNDOTBS1' I get the following error:
âORA-01548: active rollback segment '_SYSSMU1$' found, terminate dropping tablespaceâ
Here's what I have tried so far:
- Bounced the DB several times hoping to get rid of the reference to the active rollback segment. No luck.
- Created a new undo table space
- Tried 'Alter system set undo_tablespace=new_one' (noticed that 10 new rollback segs were created '_SYSMU11$..SYSMU20$' pointing to the new undo tablespace)
- I then performed an offline drop of the undo tablespace data file. After the offline drop, the system rollback segs SYSMU01$..SYSMU10$ went offline as would be expected.
- Even with the new undo tablespace, all other application tablespaces still use the old undo which due to the offline drop, no inserts, deletes, etc. will work.
After all the above, I still get the same error when trying to drop the undo tablespace.
Any ideas? Ideally I would like to be able to drop the undo tablespace and recreate it, but at this point I'd be happy to bring back the undo tablespace datafile so we can at least use the database again.
Thanks in advance!!
Colin

SQL> show parameters undo
undo_retention integer
10800
undo_suppress_errors boolean
FALSE
undo_tablespace string
UNDOTBS03
r.- The undo tablespace which the database is looking for is this one ( UNDOTBS03 )
What is strange about the lines above is that I have
all those params above commented out in my
\pfile\init.ora file (and have bounced the db).
Iâm not sure why they are still showing up.
r.- Perhaps your instance is started with the spfile instead of the init file
Here's what we did after the UNDOTBS1 tablespace
caused the disk to run out of space during a large
import (from now on we are going to use commit = y):
1. Created a new undo tablespace UNDOTBS03 - got
errors so instead we to created it as a temp
tablespace UNDOTBS03 which worked. r.- Yes, but that tablespace was created as a UNDO TABLESPACE ?
2. Tried to drop UNDOTBS1 got the following error
msg:
ORA-01548: active rollback segment '_SYSSMU1$' found,
terminate dropping tablespacer.- Of course, the database did not allow to drop it beacause there were pending transactions
3. SQL> alter system set undo_tablespace =
'UNDOTBS03' scope = both. Tried dropping UNDOTBS1.
Same error as in #2 above.r.- There were still pending transactions with the older undo tablespace
4. We then offline dropped the datafile associated
with UNDOTBS1 hoping that would then allow us to drop
UNDOTBS1. Same error as #2.r.- That was the worst you did. You can not drop a tablespace so...
5. Then tried creating a undo tablespace again:
Create UNDO tablespace undo_temp
datafile 'D:\oracle\oradata\usbperf\undo_temp.DBF'
size 25m autoextend on;
Error on line 0
Create UNDO Tablespace undo_temp
datafile 'D:\oracle\oradata\usbperf\undo_temp.
ORA-00604: error occurred at recursive SQL level 1
ORA-00376: file 2 cannot be read at this time
ORA-01110: data file 2:
'D:\ORACLE\ORADATA\USBPERF\UNDOTBS01.DBF'
So thatâs where we are. I'm guessing that the
original undo tablespace corrupted when it tried to
extend and ran of disk space?
Thanks again for the help.
Colin ok I can guide to solve this but you have to do what I am going to tell you step by step. Is this a production database ?
Joel PÃ©rez
http://www.oracle.com/technology/experts

Undo Tablespace Reuse

Hi,
I am using 10.2.0.3.0 on Oracle Solaris 5.10.
I have an undo tablespace 'UNDOTBS1' which has 5 data files of sizes 60GB. Here are my undo parameters:
SQL> show parameter undo
NAME                                 TYPE        VALUE
undo_management                      string      AUTO
undo_retention                       integer     900
undo_tablespace                      string      UNDOTBS1I have a long running batch process, which is inserting billions of records in the table and generating lots of undo. The batch is running for the last two days, and generating lots of undo and developer says that he is committing after 3 million records, though he was told to commit after 10000 records. Now all the data files of undo tablespace have become full except one, which is also getting filled up.
I see from dba_rollback_segs that there are some 25 segments which are showing status offline.
does it mean that the oracle will reuse those segments and I dont need to add another data file to the undo tablespace?
How would I know during this running process, that oracle would reuse space in undo tablespace, or will just throw an error, after exhausting all the data files.
Thanks and regards

Hi,
As the undo segments are used in a curcular fashion, dba_free_space may not give correct picture. use following sql.
col PCT_USED head '% Used'
select a.used, b.allocated, round((a.used/b.allocated)*100,2) PCT_USED, c.unexpired, d.active from
(select sum(BYTES/1024/1024) used from dba_undo_extents where status in ('UNEXPIRED','ACTIVE')) a,
(select bytes/1024/1024 allocated from dba_data_files where tablespace_name = 'UNDOTBS1') b ,
(select sum(bytes/1024/1024) unexpired from dba_undo_extents where status = 'UNEXPIRED') c,
(select sum(bytes/1024/1024) active from dba_undo_extents where status = 'ACTIVE') d
Here ACTIVE - they are being used
UNEXPIRED - they will not be released until the undo_retention time. But if all other extents are used, then oracle may reallocate unexpired blocks and any session which is trying to read from unexpired block may received "snapshot too old error".
EXPIRED - here these blocks will be reused, you can consider them as free.
Thanks,
Siva

Dropping Undo Tablespace

Hi,
I am running Oracle 9.2 on Solaris. I have Automatic undo management turned on and have a Undo tablespace by name UNDOTBS. It has a single datafile and has AUTOEXTEND feature turned on. My UNDO_RETENTION was set to 4 hours.
Presently, the size of UNDOTBS has grown to around 40G and has filled up the file system in which it resides. I need to clean up the file system. In order to do this, I followed the below procedure.
SQL> create undo tablespace UNDOTBS1 datafile '/u04/oradata/proddb/undo_00.dbf' size 1024M autoextend off;
SQ> alter system set undo_tablespace=undotbs1 scope=both;
SQL> alter system set undo_retention=3600 scope=both;
Now, i checked the alert log and it is offlining all the UNDO segments of the first undo tablespace UNDOTBS. It is taking time and has been more than 1.5 hours. I have a couple of questions here,
1. As i have issued "alter system undo_tablespace=undotbs1 scope=both", all the new transactions will use the new UNDOTBS1. No new transactions will be using the first undo tablespace UNDOTBS right?
2. Since, i have issued "alter system undo_retention=3600 scope=both", all the transactions in first undo tablespace UNDOTBS will be rolled back after 2 hours, no matter if it is commited or not. After this period, i am safe to drop the first undo tablespace using "drop tablespace UNDOTBS including contents and datafiles;". Did i get it right?
Any information will be of a great help.
THanks,
Harris.

Hi,
Thank you all for the replies. It has been two days since i switched the undo_tablespace to use the new tablespace(undotbs1) but my alert log is still being flooded with " Undo tablespace 1 moved to pending switch out state". I tried the following to see if there are any rows returne from the above query and it did 3 rows, showing some undo segments with the following output
SQL> select a.name, b.status from v$rollname a, v$rollstat b where a.name in (select segment_name from dba_segments where tablespace_name='UNDOTBS') and a.usn=b.usn;
Name Status
_SYSSMU15$      UNKNOWN
_SYSSMU33$      UNKNOWN
_SYSSMU9$        PENDING OFFLINE
I have seen gone through the metalink id: 341372.1 that speaks about this SWITCH-OUT state. How can i make sure what old transactions are still using the Undo segments in first undo tablespace and how can i go about killing them. Could anyone please let me know that i could use the following queries to find the correct sessions that are using the Old Undo tablespace.
SQL> select ADDR, XIDUSN,STATUS, START_TIME from v$transaction;
SQL> SELECT
a.usn,
a.name,
b.status,
c.tablespace_name,
d.addr,
e.sid,
e.serial#,
e.username,
e.program,
e.machine,
e.osuser
FROM
v$rollname a,
v$rollstat b,
dba_rollback_segs c,
v$transaction d,
v$session e
WHERE
a.usn=b.usn AND
a.name=c.segment_name AND
a.usn=d.xidusn AND
d.addr=e.taddr AND
b.status='PENDING OFFLINE';
Just wanted to make sure that the second query lists the PENDING OFFLINE segments of the first tablespace right?
This is a prod. box, so wanted to confirm with you guys before proceeding.
Thanks,
Harris.

ORA-30036: unable to extend segment by 8 in undo tablespace 'UNDOTBS1'

Hi experts,
I'm getting this error
"ORA-30036: unable to extend segment by 8 in undo tablespace 'UNDOTBS1'"
will you please tell me the reason and suggest me, what to do.
OS - SunOS 5.10
oracle - 10.2.0.3.0

Hallo.
I have this problem 30036 after this request (ver. 10.2):
DELETE FROM AUDITDBA.SEOSDATA where COMPUTERNAME ='xxxxx' AND TIMSTAMP >= '18-03-2011' (! number of rows is less as 2 000 000)
I tried do this:
ALTER SYSTEM SET UNDO_RETENTION = 5;
But after next action "DELETE" again problem 30036..
Here is a "space status" from the system:
TABLESPACE_NAME VOLNE_MIESTO
CA_AUDIT 3714,1875 MB
SYSAUX 42,1875 MB
SYSTEM 5,25 MB
UNDO_TEMP 7998,0625 MB
USERS 999,125 MB
It's clear for me that datafile UNDOTBS is small but I have same questions:
1. What You recommend - use autoextend on for the UNDOTBS datafile or extend it manualy?
2. What size You recommend for the UNDOTBS when I have DB size cca 350 GB, into DB incomming cca 50 events (rows) per second and I need remove from this DB aproximetly 1/2 numbers of rows (old data)
3. Recommend me best way how I can remove old data from DB without problems as "30036"...
Thank You in advance
Peter

After transaction completed in undo tablespace what would happen? will they

after transaction completed in undo tablespace what would happen?
will they take out of undo tablespace or segment?
please elaborate, thank you

Hi,
Even, if the transaction is committed, there may remains a need to retain the undo data. For instance, for the long running queries or for the flashback features. In automatic undo management, you may specify the undo retention period, which is the minimum amount of time that Oracle Database attempts to retain old undo information before overwriting it.
One may also enable the retention guarantee, which ensures that the undo data will remain in the undo tablespace, even if that means that the transaction would fail due to lack of space in the undo tablespace.
In addition, also read the following:
http://download.oracle.com/docs/cd/B19306_01/server.102/b14231/undo.htm
regards

How the storage will be allocated on Undo tablespace made of datafiles?

I have a UNDO tablespace with two data files. Wherein one data file with Auto Extend as Yes and other as No.
In the case of multiple datafiles in UNDO tablespace,
1. Will the storage be allocated of same size on each datafile ?
2. What will be the case when the datafile with Auto Extent as NO reached it's max size?
3. Will the storage be allocated proportion to the datafile size in each file?

But when I use the following query, its not getting fired. Its taking "INDEX FULL SCAN"I would have thought that the later part of the above statement contradicts with the earlier part of the statement.
its not getting firedcontradicts this statement:
Its taking "INDEX FULL SCAN"Since the birth_dt is not the leading column of your index ...
Is it doing a INDEX FULL SCAN on the GPM_PERSONPROFILEDEMO_DOB2_IDX index?
if you have queries that involve profile_update_dt + birth_dt and some others that involve only birth_dt, then why not have the index in birth_dt + profile_update_dt instead?

Undo Tablespace Datafile is corrupted

Here's my case:
- our Oracle 9i R2 database is running in NOARCHIVE mode in Windows 2000 plateform.
- the Undo Tablespace datafile corrupted in the filesystem level (it gives "cyclic redundancy check" error when trying to copy or read from the file). Therefore database cannot startup.
- There is no backup files.
- I tried to create an Undo Tablespace to replace the existing one in the MOUNT state. That was not of course possible.
- I tried to use ALTER DATABASE CREATE DATAFILE '' AS '' statement, but it couldn't work because the original file is corrupted.
How can I solve the problem and make the database startup?
Thanks in advance.

This is a very bad situation, here's a statement from an Oracle note:
If the rollback segment datafile is physically missing, has been offlined
dropped, or the rollback segment header itself is corrupt, there is no way to
dump the transaction table to check for active transactions. So the only thing
to do is get the database open, export and rebuild. Log an iTar to engage support
to help with this process.
If you cannot get the database open, there is no other alternative than
restoring from a backup. ...
So you should contact Oracle support.

If I have two undo tablespace, under what situation oracle will use another

if I have two undo tablespaces, under what situation oracle will use another
one is default undo tablespace, another is created manually
will it use it when the default one is full?
it seem like to be it won't use the second undo tablespace.
thanks
Edited by: user11402556 on Jul 3, 2010 2:18 PM

You can have two Undo Tablespaces created with the CREATE UNDO TABLESPACE syntax.
However, the current database instance will use the Undo Tablespace specified by the instance level parameter undo_tablespace.
When you want to switch the datafiles of an undo tablespace from, say, one mount point to another (or when you want to "resize" an undo tablespace that has grown very large for specific one-off / adhoc jobs), you would create another undo tablespace, reset the instance level parameter to the new undo tablespace and wait for a decent interval (e.g. at least undo_retention or autotuned undo retention) before dropping the old undo tablespace.
So yours is not a hypothetical question. Having two undo tablespaces is a real-world practice -- but the two are concurrently present for only a short duration, while transactions and queries switch from one to the other.
(In RAC, each database instance must have it's own Undo tablespace as well).
Hemant K Chitale

Undo tablespace with active transaction/segment corrupted

How to recover if undo tablespace with active transaction/segment corrupted

Try RMAN blockrecover or restore the tablespace from latest backup and recover it.

Create undo tablespace while at mount state, possible?

Oracle 10g (10.2.0.1.0), on RedHat 4 (both 64 bit)
My test db’s undo tablespace is corrupt (no RMAN or user backups).
DB fails to open, fails during "Doing block recovery for file 2 block 209910"Is there a way of creating undo tablespace while at mount state?
I am pretty confident it’s not possible, also searched the net and found no fix, I thought I will ask.
I get error:
CREATE UNDO TABLESPACE UNDOTBS1_new DATAFILE '/u01/oracle/oradata/TEST_DB1/undotbs2_01.dbf' SIZE 512M
ERROR at line 1:
ORA-01109: database not openp{code}As a last option, I will open the db with *'reset logs'*, wondering is any thing else I can try to fix the issue.
Thanks for any input.

your problem is not to create un UNDO, your problem is that your UNDO is corrupted ! ;)
You must recover undo because it may contains needed data.
As there is not backup, you may try to set undo to manual, and then open the db and recreate the undo
but this may fail and you may have to allow undo corruption in initialization (_corrupted_rollback_segments), but this is an unsupported operation (hidden parameter)
Could you contact support ?
is your db a production one ?
Edited by: Leo Anderson on 1 sept. 2008 15:25

Cannot drop undo tablespace with Active extents but no active sessions

Hi all,
I am running on 10gr2 oracle database in a Linux 32 bit environment.
I have a corrupted undo datafile. I've been able to recover the datafile but still the alert logs is still showing that it is corrupted.
Now what I'm trying to do is to drop the old undo tablespace like the one discussed here Change undo tablespace in Oracle - Switch Oracle's Undo tablespace.
My problem is I have an Active extent but no active session corresponding to it:
SYS@ccasdb AS SYSDBA > SELECT a.name,b.status
FROM   v$rollname a,v$rollstat b
WHERE a.usn = b.usn
AND    a.name IN (
                  SELECT segment_name
                  FROM dba_segments
                  WHERE tablespace_name = 'UNDOTBS1'
                 ); 2    3    4    5    6    7    8
NAME                           STATUS
_SYSSMU18$                     PENDING OFFLINE
SYS@ccasdb AS SYSDBA > SELECT a.name,b.status , d.username , d.sid , d.serial#
2 FROM   v$rollname a,v$rollstat b, v$transaction c , v$session d
3 WHERE a.usn = b.usn
4 AND    a.usn = c.xidusn
5 AND    c.ses_addr = d.saddr
6 AND    a.name IN (
7               SELECT segment_name
8               FROM dba_segments
9               WHERE tablespace_name = 'UNDOTBS1'
10              );
no rows selected
but still cannot drop the undotablespace.
SYS@ccasdb AS SYSDBA > drop tablespace UNDOTBS1 including contents and datafiles;
drop tablespace UNDOTBS1 including contents and datafiles
ERROR at line 1:
ORA-30013: undo tablespace 'UNDOTBS1' is currently in use
Regards,
Tim

Yes i was able to recover the datafile but the alert_log kept on prompting that it is still corrupted. The datafile's status is online.
I want to kill the session/s contributing to the PENDING OFFLINE status of the rollback segment so I can already drop the undotablespace and so
that the alert log will stop prompting for that corrupted datafile in that UNDOTABLESPACE.
Can i Just alter the tablespace's datafiles to offline drop then just drop it. Then I drop the tablespace?

Unable to shrink undo tablespace... Help!

Hi,
I have problems to shrink the system undo tablespace, which has grown up to 14 GB.
I use 9.2. Table space owner is 'SYSTEM', undo_management = AUTO.
I tried to shrink the greatest rollback segments by the commands
ALTER SESSION SET UNDO_SUPPRESS_ERRORS = TRUE;
ALTER ROLLBACK SEGMENT "_SYSSMU6$" SHRINK TO 20 M;
Oracle confirmed these commands, but nothing happened.
What am I doing wrong?
Hermann Mueller

You have seen the discussion about the undo segments, on the temporary tablespaces, you should be aware that the sort segment of a given temporary tablespace is created at the time the first sort operation takes place. The sort segment continues to grow by means of extent allocation until the segment size has reached the total storage demands of all of the active sorts running on the instance. Oracle will keep on allocating temporary space on demand unless the physical limit states otherwise.
Temporary segments are produced each time a sort operation (explicit -order by- or implicit -aggregation, reindexing-) requires to sort a set that cannot fit into memory. So if you detect excesive sort usage you should aim your monitors towards the sort operations (reports, reindexing, max, min, aggregations, order by ...). If your system has a DDS behaviour, this kind of operations are frequent as a massive sorting has to be peformed against millions of rows.
A Temporary tablespace will almost always appear to be near 100% full, that's because once oracle has allocated temporary space it doesn't release it back to the free space, it keeps it allocated even when the sort operation has finished. Criteria behind this fact is similar to the one oracle used to have when the rollback segments were in use, Oracle only allocated space and the dba should perform manual actions to release space, and the criteria is performance. Once it has allocated space this big, there are possibilities that the same circumstances that raised the temporary usage high water mark to this level are repeated, so if oracle keeps the mamimum allocated space, it won't have to allocate the same storage once more.
Main concern with temporary tablespace growth is not free space itself, but the reasons why this amount of space was allocated, so I suggest you to track the sql statements with sort operations. If you are certain the circumstances that motivated this amount of temporary resources to be allocated won't be repeated again, then you could think of resizing down your temporary tablespace. I suggest you to create a new temporary tablespace with the desired size, and alter the default tempoary tablespace to point to this newly crated temporary tablespace, and finally get rid of the original temporary TS.
~ Madrid

Segment shrinking and UNDO tablespace

When i issued
alter table <table_name> shrink space cascade;
I got the error 'Unable to extend UNDO ...by 8' . Does Segment shrinking consume space from UNDO tablespace?
Message was edited by:
for_good_reason

As Jonathan said, shrinking segment generates redo and undo data.
But, this phoenomenon of undo shortage is not normal case.
Shrinking segment might invole continous DML(deleting and re-inserting),
but it's transaction seems to be committed internally at regular intervals.
For this reason, shrinking operation should not hold undo area that long.
But have no knowledge on exact behavior of shrink operation.
Someone else will shed a light.
How big your table and indexes?
Some cases are reorted that shrinkage on big segment generates really large amount of undo data.
This might be related with your problem. But not sure.
Visit metalink note# 3888229.

Undo tablespace corruption will terminate DB operation

Similar Messages

Maybe you are looking for