Email: service@parnassusdata.com 7 x 24 online support!

Oracle untruncate using PRM-DUL case study

Oracle untruncate using PRM-DUL case study

Case Study on Oracle database recovery via PRM-DUL

 

 

CASE 1: Truncate table by mistake

User D had truncated a table by mistake on production environment. The DBA tried to recover table from RMAN backup, and accidently the backup is unavailable. Therefore DBA decided to use PRM-DUL for rescuing all truncated data.

Since all database system files are healthy, DBA just needs to load SYSTEM table data file in dictionary mode and TRUNCATED table file. For example:

 

create table ParnassusData.torderdetail_his1  tablespace  users asselect * from parnassusdata.torderdetail_his; 

 

SQL> desc  ParnassusData.TORDERDETAIL_HIS
 Name                    Null?    Type
 ———————– ——– ————–
SEQ_ID                NOT NULL   NUMBER(10)
 SI_STATUS                        NUMBER(38)
 D_CREATEDATE                     CHAR(20)
 D_UPDATEDATE                     CHAR(20)
 B_ISDELETE                       CHAR(1)
 N_SHOPID                         NUMBER(10)
 N_ORDERID                        NUMBER(10)
 C_ORDERCODE                      CHAR(20)
 N_MEMBERID                       NUMBER(10)
 N_SKUID                          NUMBER(10)
 C_PROMOTION                      NVARCHAR2(5)
 N_AMOUNT                         NUMBER(7,2)
 N_UNITPRICE                      NUMBER(7,2)
 N_UNITSELLINGPRICE               NUMBER(7,2)
 N_QTY                            NUMBER(7,2)
 N_QTYFREE                        NUMBER(7,2)
 N_POINTSGET                      NUMBER(7,2)
 N_OPERATOR                       NUMBER(10)
 C_TIMESTAMP                      VARCHAR2(20)
 H_SEQID                          NUMBER(10)
 N_RETQTY                         NUMBER(7,2)
 N_QTYPOS                         NUMBER(7,2)select count(*) from ParnassusData.TORDERDETAIL_HIS;COUNT(*)

 

———-

984359

 

select bytes/1024/1024 from dba_segments where segment_name=’TORDERDETAIL_HIS’ and owner=’PARNASSUSDATA';

 

BYTES/1024/1024

—————

189.71875

 

 

 

SQL> truncate table ParnassusData.TORDERDETAIL_HIS;

 

Table truncated.

 

SQL> select count(*) from ParnassusData.TORDERDETAIL_HIS;

 

COUNT(*)

———-

0

 

 

 

Run PRM-DUL, and select Tools =>Recovery Wizard

 

PRM-DUL-DUL12

 

Click Next

 

PRM-DUL-DUL13

 

Client did not user ASM storage, therefore just select ‘Dictionary Mode’:

PRM-DUL-DUL14

 

Next, we need to select some characters: including Endian bit order and DBNAME

Since Oracle datafiles have different Endian bit orders on different OS, please choose accordingly:

 

Solaris[tm] OE (32-bit) Big
Solaris[tm] OE (64-bit) Big
Microsoft Windows IA (32-bit) Little
Linux IA (32-bit) Little
AIX-Based Systems (64-bit) Big
HP-UX (64-bit) Big
HP Tru64 UNIX Little
HP-UX IA (64-bit) Big
Linux IA (64-bit) Little
HP Open VMS Little
Microsoft Windows IA (64-bit) Little
IBM zSeries Based Linux Big
Linux x86 64-bit Little
Apple Mac OS Big
Microsoft Windows x86 64-bit Little
Solaris Operating System (x86) Little
IBM Power Based Linux Big
HP IA Open VMS Little
Solaris Operating System (x86-64) Little
Apple Mac OS (x86-64) Little

 

In traditional UNIX, AIX (64-bit), UP-UNIX (64-bit), it use Big Endian bit order,

 

PRM-DUL-DUL15

 

Usually, Linux X86/64, Windows remain default Little Endian:

PRM-DUL-DUL16

 

Attention: if your data file was generated on AIX, if you want to recover data on window, please select original Big Endian format.

Since the data file is on Linux X86, we select Little as Endian, and input database name. (The input database name can be different from DB_NAME found in datafile header,  the input database name is just an alias. PRM-DUL will check if your PRM-DUL license is valid , the valid license key is generated based on DB_NAME found in datafile header)

PRM-DUL-DUL17

 

Click Next =>Click Choose Files

 

Usually, if the database is not too big, we could select all data files together; if the database capacity is huge and DBA knows the data location, at least you should  select both SYSTEM tablespace and specified datafile.

 

Attention, the GUI Supports Ctrl + A & Shift short keys:

 

PRM-DUL-DUL18

 

PRM-DUL-DUL19

 

Specify the Block Size (Oracle data block size) according to the real circumstance. For example, if default DB_BLOCK_SIZE is 8K, but part of tablespaces’ block size is 16k,then user has to specify them as correct block size one by one.

 

OFFSET setting are just for raw device storage mode, for example: on AIX, based on LV of  normal VG, the offset will be 4k OFFSET.

 

If you are using raw device but don’t know what the OFFSET is, please use dbfsize tool which is under $ORACLE_HOME/bin

 

 

$dbfsize /dev/lv_control_01Database file: /dev/lv_control_01Database file type: raw device without 4K starting offsetDatabase file size: 334 16384 byte blocks

 

Since all data file block size here is 8K and there is no OFFSET, please click load:

 

PRM-DUL-DUL20

 

PRM-DUL read Oracle dictionary directly, and recreate a new dictionary in embedded database. It can help us to recuse most types of data in Oracle DB.

 

PRM-DUL-DUL21

 

After recreating dictionary, the dialog show character information:

PRM-DUL-DUL22

 

Attention: PRM-DUL supports multiple languages and multiple Oracle character set. However, the prerequisite is the OS had installed specified language packages. For example, on Windows, if you didn’t install Chinese language package, even Oracle database characters are independent and support ZHS16GBK, PRM-DUL would display Chinese as messy code. Once the Chinese language package is installed on OS, PRM-DUL can display multibyte character set properly.

 

Similarly, on Linux, it need font-Chinese language package.

 

[oracle@mlab2 log]$ rpm -qa|grep chinesefonts-chinese-3.02-12.el5

 

After loading, in PRM-DUL GUI, it displayed database tree diagram by database users.

 

Click Users, you can find more users, for example, if user want to recover a table under PARNASSUSDATA SCHEMA, click PARNASSUSDATA, and double click that table:

 

PRM-DUL-DUL23

 

Previously TORDERDETAIL_HIS had been truncated, so it  won’t show any data . Please select unload truncated Data:

 

PRM-DUL-DUL24

 

PRM-DUL will scan the tablespace and extract data from truncated table.

PRM-DUL-DUL25

 

PRM-DUL-DUL26

 

As in the above picture, the truncated TORDERDETAIL_HIS had exported 984359 record, and saved to specified falt file.

In addition, it generated SQLLDR control file for text data importing

 

$ cd /home/oracle/PRM-DUL/PRM-DULdata/parnassus_dbinfo_PARNASSUSDATA/$ ls -l ParnassusData*-rw-r–r– 1 oracle oinstall       495 Jan 18 08:31 ParnassusData.torderdetail_his.ctl-rw-r–r– 1 oracle oinstall 191164826 Jan 18 08:32 ParnassusData.torderdetail_his.dat.truncated 

 

$ cat ParnassusData.torderdetail_his.ctl

LOAD DATA

INFILE  ‘ParnassusData.torderdetail_his.dat.truncated’

APPEND

INTO TABLE ParnassusData.torderdetail_his

FIELDS TERMINATED BY ‘ ‘

OPTIONALLY ENCLOSED BY ‘”‘

TRAILING NULLCOLS (

“SEQ_ID” ,

“SI_STATUS” ,

“D_CREATEDATE” ,

“D_UPDATEDATE” ,

“B_ISDELETE” ,

“N_SHOPID” ,

“N_ORDERID” ,

“C_ORDERCODE” ,

“N_MEMBERID” ,

“N_SKUID” ,

“C_PROMOTION” ,

“N_AMOUNT” ,

“N_UNITPRICE” ,

“N_UNITSELLINGPRICE” ,

“N_QTY” ,

“N_QTYFREE” ,

“N_POINTSGET” ,

“N_OPERATOR” ,

“C_TIMESTAMP” ,

“H_SEQID” ,

“N_RETQTY” ,

“N_QTYPOS”

)

 

 

When you import data to original table, ParnassusData strongly recommends you to modify SQLLDR table name as a temp table, it would not impact your previous environment.

 

 

 

 

$ sqlldr control=ParnassusData.torderdetail_his.ctl direct=yUsername:/ as sysdba//user SQLLDR to import data//Minus can be used for data comparing

 

select * from ParnassusData.torderdetail_his minus select * from parnassus.torderdetail_his;

 

no rows selected

 

 

 

After diffing, there is no difference between original data and PRM-DUL exported data.

PRM-DUL successfully recovered the truncated table

 

 

CASE 2: Recovery mis-truncated table by DataBridge

 

In Case 1, we use traditional unload+sqlldr for data recovery, but actually ParnassusData would like to strongly recommend using  DataBridge Feature for recovering.

 

Why use DataBridge?

 

  • Traditional unload+sqlldr means a copy of data needs to be saved as flat file on filesystem first, data has to be loaded into Unicode text file and then inserted into destination database by sqlldr, this will take double storage and double time.
  • DataBridge can extract data from source DB and export to destination DB without any intermediary.
  • Once the data arrived destination DB, user can begin to validate them.
  • If source and destination database located on different servers, then read/write IO will be balanced on two servers , MTTR will be saved.
  • If DataBridge is used in truncated table recovery, it is very convenient that truncated data can be exported back to problem database directly.

 

DataBridge is very simple and convenient. Right click the table on the left side, and select DataBridge:

 

PRM-DUL-DUL27

 

 

As the first time to use DataBridge, DB connection information is necessary, which is similar with SQL Developer connection, including: DB host, Port, Service_Name and Account information.

Attention: DataBridge will save data to the specified schema given in the DB connection.

 

PRM-DUL-DUL28

 

AS above G10R25 connection, user is maclean, and the corresponding Oracle Easy Connection is

192.168.1.191:1521/G10R25。

 

After inputting the account/connection information, you can use test for connection testing. If return message is “ Connect to DB server successfully “, the connection is done and click to save.

 

PRM-DUL-DUL29

 

After saving connection and go to DataBridge window, please select Connection G10R25 at the drop down list.

PRM-DUL-DUL30

 

If your DB connection is not in the drop down list, please click DB connection Button, which is highlighted in red.

PRM-DUL-DUL31

 

After selecting DB Connection, Tablespace dropdown list will be selectable:

PRM-DUL-DUL32

 

 

Attention on DataBridge recovering truncated/dropped table: when you recovering truncated/dropped and insert data back to source DB, users should choose another tablespace which diffs from the original tablespace. If export data into same tablespace, oracle will reuse space which stores truncated/dropped table, and can make data overwritten, we will lose the last resort to recover the data.

 

For example, we truncated a table and would like to user DataBridge to recover data back to source database, but we would like to use another table name. Original table name is torderdetail_his, and user can select “if need to remap table” and input proper destination name, as below:

PRM-DUL-DUL33

 

 

 

Attention: 1) For destination DB which already had the same table name, PRM-DUL will not recreate a table but append all recovered data. 2) For destination DB which did not have source table name, PRM-DUL would try to create table and recover the data.

 

In this case, we would recover Truncated data, therefore, please select “if data truncated?” checkbox, Or, PRM-DUL would do regular data extraction, but not Truncated data.

 

Truncate recovery methodology is: Oracle will only update table DATA_OBJECT_ID in data dictionary and segment header. Therefore, the real data will not be overwritten. Due to the difference between dictionary and DATA_OBJECT_ID, Oracle server process will not read truncated data while scanning table. But, the real data is still there.

 

PRM-DUL will try to scan 10M-bytes blocks which are behind of the table’s segment header, if some blocks with smaller DATA_OBJECT_ID than the object’s current DATA_OBJECT_ID, then PRM-DUL thinks it find something useful.

 

 

There is a blank input field called ”if to specify data object id”, which let user input Data Object ID. Usually, you don’t need to input any value, unless the recovery does not work. We suggest user to contact ParnassusData for help.

 

Click DataBridge button ,then it will start extracting if the configuration is done.

 

PRM-DUL-DUL34

 

DataBridge will display the successfully rescued rows and elapsed time.

 

PRM-DUL-DUL35