分类目录归档：Oracle

在oracle恢复中ORA-600 4194是一个非常常见的错误,该错误的主要原因是由于redo记录和undo(rollback)记录不匹配.
ORA 600 4194错误原因以及含义

ERROR:              

  Format: ORA-600 [4194] [a] [b]
 
VERSIONS:           
  versions 6.0 to 12.1 

DESCRIPTION:

  A mismatch has been detected between Redo records and rollback (Undo) 
  records.

  We are validating the Undo record number relating to the change being 
  applied against the maximum undo record number recorded in the undo block.

  This error is reported when the validation fails.

ARGUMENTS:
  Arg [a] Maximum Undo record number in Undo block
  Arg [b] Undo record number from Redo block

ORA 600 4194 错误处理思路
第一步

Confirm whether the database is up and running or not.  If the database fails to start or crashes shortly 
after startup due to this error occurring, then try setting event 10513 at level 2 in the init.ora/spfile 
to disable transaction recovery and restart the instance, e.g.:

      event = "10513 trace name context forever, level 2"

This may allow the database to successfully open and stay up so that 
the required diagnostics/actions can be performed.

第二步

In the trace file there should be an undo segment header dump, and so check 
to see if the undo segment header shows an active transaction after recovery, e.g.:


TRN TBL    <---- Represents the Transaction table for the particular undo segment

index state cflags wrap# uel scn dba
--------------------------------------------------------------------------------------------- 
0x41 9 0x80 0x35ab6 0xffff 0x0695.38f6b959 0x1081e796 
0x42 9 0x80 0x35bb1 0x000e 0x0695.38f6b028 0x1081e793 
0x43 9 0x80 0x35b11 0x005d 0x0695.38f6b7ae 0x1081e795 
0x44 9 0x80 0x359f0 0x0036 0x0695.38f69a91 0x1081e78e 
0x45 10 0x80 0x35b1b 0x0000 0x0695.3a0aba4d 0x1081e796 
0x46 9 0x80 0x35bb7 0x001c 0x0695.38f69bde 0x1081e78f 
===================================

State ---> This column specifies the status of the transaction 
                  9 -----> represents a commited transaction
                  10 ---> Represents a active transaction
Dba -----> Undo block containing the undo records
                  Strictly speaking this is the block at the end of the undo chain.

You can see from the transaction table that there is an active transaction 
for this particular rollback/undo segment after recovery.
Therefore this rollback/undo segment and/or undo tablespace cannot be dropped without corrupting the database!  
Therefore recreating the UNDO tablespace is not an option.

第三步

From the trace file determine the affected undo segment, e.g.:

Block image after block recovery:

UNDO BLK: 
xid: 0x0015.02b.0001544b seq: 0x163e cnt: 0x12 irb: 0x12 icl: 0x0 flg: 0x0000

XID ==> Undo segment no + Slot no + Sequence no

Therefore, in this case the Undo Segment is:

USN# 0x15 (Hex) ==> 21 (Dec)  ==> _SYSSMU21$

So if and ONLY IF the transaction table shows no active transaction can the
 rollback/undo segment be offlined and dropped.Note however, 
that before you can confirm if the entire UNDO tablespace can be dropped, you would need to check the 
transaction tables of ALL active rollback/undo segments in the same wasy as the above.  
The steps required to drop the rollback/undo segment are fully detailed in Note:179952.1, 
but are briefly listed here for completeness:

If using Automatic Undo Management


Offline the undo segment using the _OFFLINE_ROLLBACK_SEGMENTS parameter and bounce the database as follows:

1.  Create  and edit the init.ora file for the instance to set the following parameters:

UNDO_MANAGEMENT=MANUAL 
_OFFLINE_ROLLBACK_SEGMENTS=(_SYSSMU21$)

2.  Open the database in restricted mode to prevent user access, e.g.:

connect / as sysdba
startup restrict pfile = '<Full path to init.ora file>';

3.  Drop the rollback/undo segment, e.g.:

drop rollback segment "_SYSSMU21";

4.  Shutdown the instance, and remove the init.ora parameters added in point 1 and restart the instance, e.g.:

shutdown immediate
startup


If SMON was recovering the transaction then this may not work as we cannot open the database if it is determined 
to be in an inconsistent state. I have reviewed a number of SRs where this approach was successful, 
so it is important to try it first but understand that it may fail and you will have to resort to 
a point in time recovery or forcing open the DB and recreating it.

第四步

Now we need to dump the undo block to see which object was affected.  
We noted in Step 2 that this is the active transaction (from the trace file): 

TRN TBL 

index state cflags wrap# uel scn dba 
0x45 10 0x80 0x35b1b 0x0000 0x0695.3a0aba4d 0x1081e796 

Dba----------------> Undo block containing the undo records 

dba--->0x1081e796 is the block containing the active transaction . 


Use the WebIV tools to convert this RDBA to block number (block#) and file number (file#), e.g.: 

V SPLIT ==> DBA (Hex) = File#,Block# (Hex File#,Block#) 
= ===== === ===== ============ 
V8 10,10 ==> 276948886 (0x1081e796) = 66,124822 (0x42 0x1e796) 


So the file# is 66 and the block# is 124822, so dump the block by issuing: 

SQL> Alter system dump datafile 66 block 124822; 


This will generate a trace file in the user_dump_dest.  The following is a sample of the information in the undo block:

UNDO BLK: 
xid: 0x000c.045.00035b1b seq: 0x1e14 cnt: 0x17 irb: 0x17 icl: 0x0 flg: 0x0000 

Rec Offset Rec Offset Rec Offset Rec Offset Rec Offset 
--------------------------------------------------------------------------- 
0x01 0x1f8c 0x02 0x1f30 0x03 0x1ed4 0x04 0x1e78 0x05 0x1e1c 
0x06 0x1dc0 0x07 0x1d64 0x08 0x1d08 0x09 0x1cac 0x0a 0x1c50 
0x0b 0x1bf4 0x0c 0x1b98 0x0d 0x1b3c 0x0e 0x1ae0 0x0f 0x1a74 
0x10 0x1a18 0x11 0x19bc 0x12 0x1960 0x13 0x1904 0x14 0x187c 
0x15 0x181c 0x16 0x1798 0x17 0x173c 

* Rec #0x16 slt: 0x45 objn: 1485619(0x0016ab33) objd: 1485619 tblspc: 71(0x00000047) 
* Layer: 11 (Row) opc: 1 rci 0x00 
Undo type: Regular undo Begin trans Last buffer split: No 
Temp Object: No 
Tablespace Undo: No 
rdba: 0x00000000 
*----------------------------- 
uba: 0x1081e796.1e14.14 ctl max scn: 0x0695.38f69853 prv tx scn: 0x0695.38f698a1 
KDO undo record: 
KTB Redo 
op: 0x04 ver: 0x01 
op: L itl: scn: 0x0019.009.00034237 uba: 0x36c0cce4.1d2f.19 
flg: C--- lkc: 0 scn: 0x0695.38f6b96b 
KDO Op code: URP xtype: XA bdba: 0x35406893 hdba: 0x35406892 
itli: 1 ispac: 0 maxfr: 4863 
tabn: 0 slot: 0(0x0) flag: 0x2c lock: 0 ckix: 0 
ncol: 1 nnew: 1 size: -1 
col 0: [ 4] c3 0e 36 2e 
*----------------------------- 

* Rec #0x17 slt: 0x45 objn: 1485619(0x0016ab33) objd: 1485619 tblspc: 71(0x00000047) 
* Layer: 11 (Row) opc: 1 rci 0x16 
Undo type: Regular undo Last buffer split: No 
Temp Object: No 
Tablespace Undo: No 
rdba: 0x00000000 
*----------------------------- 

From the trace file above:

UNDO BLK: 
xid: 0x000c.045.00035b1b seq: 0x1e14 cnt: 0x17 irb: 0x17 icl: 0x0 flg: 0x0000

The undo segment with the active transaction is segment is 0x000c (Hex) which is 12 (Dec) as the XID is:

      Undo segment no + Slot no + Sequence no

This step is often skipped because it was performed earlier in step 3, however it is a good idea to do this 
again now to make sure that the XID from the UNDO block matches the UNDO SEGMENT HEADER, 
this way you have followed all the chain, from the UNDO SEGMENT to UNDO BLOCK, back and forth.  
If there is a conflict here please check and make sure that the customer dumped the correct undo block.

Check for the value of irb which is an index which points you to the latest change done to the undo block.
This is the point from which a rollback would begin if one was issued.

From the trace file we see: 'irb: 0x17' so this points to record 0x17, 
so search for this particular string i.e 0x17 and it will take you to undo record 'REC #0x17', e.g.:

* Rec #0x17 slt: 0x45 objn: 1485619(0x0016ab33) objd: 1485619 tblspc: 71(0x00000047) 
* Layer: 11 (Row) opc: 1 rci 0x16 
Undo type: Regular undo Last buffer split: No 
Temp Object: No 
Tablespace Undo: No 
rdba: 0x00000000 
*----------------------------- 

Note the slot number (slt) is 0x45, the object number (objn) is the OBJECT_ID from dba_objects 
and data object number (objd) is the DATA_OBJECT_ID from dba_objects.  
These numbers may be the same but not necessarily, and so if the database is open then identify this object, e.g.:

        select object_name, owner, object_type, data_object_id from dba_objects where object_id = <objn>;

This is the object, which has an active transaction.  Note in the above trace file extract that rci 
has a value of 0x16 which means that this record is at the end of an undo chain.  
This means that the chain continues in another UNDO BLOCK.  
Please refer to unpublished Note:281504.1 for information on Undo chains.

So the next record that needs to be rolled back is present in REC #X016.  
If rci is 0x00 then it means that this is the first record present in the undo chain 
and so you can check to see if there is rdba info, e.g.:


* Rec #0x16 slt: 0x45 objn: 1485619(0x0016ab33) objd: 1485619 tblspc: 71(0x00000047) 
* Layer: 11 (Row) opc: 1 rci 0x00 
Undo type: Regular undo Begin trans Last buffer split: No 
Temp Object: No 
Tablespace Undo: No 
rdba: 0x00000000 
*----------------------------- 
uba: 0x1081e796.1e14.14 ctl max scn: 0x0695.38f69853 prv tx scn: 0x0695.38f698a1 
KDO undo record: 
KTB Redo 
op: 0x04 ver: 0x01 
op: L itl: scn: 0x0019.009.00034237 uba: 0x36c0cce4.1d2f.19 
flg: C--- lkc: 0 scn: 0x0695.38f6b96b 
KDO Op code: URP xtype: XA bdba: 0x35406893 hdba: 0x35406892 
itli: 1 ispac: 0 maxfr: 4863 
tabn: 0 slot: 0(0x0) flag: 0x2c lock: 0 ckix: 0 
ncol: 1 nnew: 1 size: -1 
col 0: [ 4] c3 0e 36 2e 
*----------------------------- 


If the object is an Index, drop and recreate it.  If it is a table, 
then again the table would need to be dropped and recreated (or truncated) 
so that its object number changes and hence the rollback/undo is no longer required.  
If this isn't possible, then you have two options:

First take a backup of the database in its current state.  
This is critical in case anything goes wrong and you lose the opportunity to salvage the data! 

Option 1

 - Restore the undo segment datafile and the datafile containing the object and perform a full recovery.
   This can only be done if you have all the archived redo as you will need to do full recovery on these files.

OR 

Option 2

If option 1 is not possible, you can use the unsupported method, e.g.:

Specify the undo segment in the _OFFLINE_ROLLBACK_SEGMENTS parameter and try to drop the rollback segment.
If there is an active transaction then this is not likely to work and you will probably need 
to set the _CORRUPTED_ROLLBACK_SEGMENTS parameter as well

温馨提示：
1.隐含参数_OFFLINE_ROLLBACK_SEGMENTS/_CORRUPTED_ROLLBACK_SEGMENTS属于Oracle内部隐含参数,建议在Oracle support认可的情况下使用,因为使用之后可能导致数据库事务完整性彻底损坏
2.进行屏蔽事务之前,如果条件允许建议使用txchecker检查
2.使用上述方法恢复数据库之后，建议通过逻辑方式导出导入重建数据库