Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: Deciphering Oracle REDO Log

Re: Deciphering Oracle REDO Log

From: Wim Coekaerts <wcoekaer_at_pacbell.net>
Date: 1998/01/14
Message-ID: <34BDAF5F.36F3@pacbell.net>#1/1

mike.stevens wrote:
> I am interested in writing a program that generates the original SQL
> statements from the REDO log. The REDO log appears to be in some sort of
> binary format. Does anyone know if it's just compressed or if there is a
> way to decipher it? Any help would be greatly appreciated.
Redolog is indeed in a binary format, it contains the atomic changes to the database, made up of change vectors. Basically it is a set of opcodes. like 10.x is index change, 5.x is rollback data 11.x is table data change etc. They also only contain the changes in the blocks. Not full updates. Blockcleanouts are mentioned and what not. ALL changes to the physical structure of the database (and thus also logical representation)... are logged. Changes to the datadictionary because of space management or rollback/undo data or index changes or tabledata changes or what not are all recorded this way.

It is not really possible to write a redolog analyzer that extracts sql commands. You can write a filter that says 'these blocks have changed and all' but you cannot say, update xxx where yyy =zzz... has been executed. dml or so is not logged in the redolog files. Before changes are made to the blocks, this redo is generated, and then the redo is applied to the block. There are tools that can say which blocks have been changed when and how.

Thats all there is to it really.
Cheers
Wim Received on Wed Jan 14 1998 - 00:00:00 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US