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: log file format

Re: log file format

From: Spencer <spencerp_at_swbell.net>
Date: Fri, 1 Jun 2001 23:13:49 -0500
Message-ID: <sIZR6.486$yt.8489@nnrp1.sbc.net>

"Jonathan Lewis" <jonathan_at_jlcomp.demon.co.uk> wrote in message news:991299291.2702.0.nnrp-07.9e984b29_at_news.demon.co.uk...
> Comments in-line
>
> --
> Jonathan Lewis
> Yet another Oracle-related web site: http://www.jlcomp.demon.co.uk
>
> Practical Oracle 8i: Building Efficient Databases
> Publishers: Addison-Wesley
>
> Reviews at: http://www.jlcomp.demon.co.uk/book_rev.html
>
>
> John wrote in message <8bc78dd8.0105280541.4c2b8b56_at_posting.google.com>...
> >
> >OK
> >so chained rows are not logged?
> >
>
>
> Yes, they are logged, but again the structure
> of the logfile content probably does not lend itself
> to the simple reverse engineering that 'ordinary'
> updates allow.
>
>
> >> You would be hard pushed to demonstrate
> >> that the three operations were a single partitioned
> >> row update.
> >>
> >
> >I believe it is possible using the SCN cause they will have the same.
> >But it's hard and time-consuming.
>
>
> Which is probably why it is not done in this
> release of log-miner. Future releases are supposed
> to be much cleverer, e.g. converting:
> delete from tab$ where obj# = 12345
> into
> drop table FRED;
>
>
>
> >> alter system dump logfile '{fully qualified log file name}';
> >>
> >
> >is this file explained in the oracle documentation?
> >
>
>
> No - you hvae to figure it all out by trial and error
> for the type of data you are interested in. A couple
> of key points:
> a) You get two log items per data change, the
> rollback of the change, and the change
> b) As transactions start you get information about
> rollback segment header changes adding to the volume
> c) When a change is the first change to a block for a
> transaction, it tends to introduce logging information
> about data block header (specificially ITL) changes.
> d) Even when nothing else appears to be going on in
> the system, you get timeouts occuring which throw
> little surprise changes into the log to confuse the issue.
>

and once you get it figured out, oracle is going to desupport the release you've got it working with, and issue a new release that throws a grease covered monkey wrench into the works.

while it would be an interesting problem to work on, i'm leaving this one to the oracle folks that are tinkering with the code that writes and reads the redo logs.

if you after simply after DML changes (inserts, updates, deletes) and some simple DDL (truncates, alter table ... add column), for replication purposes, i do know of one software vendor (other than oracle) that's got code that can read the redo logs, figure out the chained row mess, and replicate the changes to other oracle databases.

email me directly if you have any questions. Received on Fri Jun 01 2001 - 23:13:49 CDT

Original text of this message

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