Re: logmnr problems

<jdarrah_co_at_my-deja.com> wrote in message news:8td3kh$ht3$1_at_nnrp1.deja.com...
> I am having problems using the dbms_logmnr package on Oracle 8.1.6.1
> running on linux. The problem is that when I query the
> v$logmnr_contents view, I see the following:
>
> update UNKNOWN.Objn:61 set Col[2] = HEXTORAW('c102'), Col[3] = HEXTORAW
> ('c102')
>
> I used the logmnr_d.build procedure to build the dictionary, but it
> still puts UNKNOWN instead of the object name in the SQL_REDO column.
> I'm guessing I'm doing something stupid here since I've had identical
> results on Solaris and I've heard about people using logminer with
> great success. Any help would be greatly appreciated as this is
> driving me crazy.
>

Two thoughts. Any segment that is created *after* you've built the dictionary file will always appear in this bafflingly-obscure format. The function of the dictionary file is to provide a conversion between the native hex format of the change vectors in the Redo files to the plain English segments we mere humans use -hence if anything is missing in the dictionary file, the conversion can't take place, and logminer reverts to the native Hex formnat.

Secondly, there is a whole bunch of objects in the data dictionary which appear simply *never* get into the dictionary file. I can't be more specific than that, I'm afraid, since my favourites, UET$ and FET$ do indeed make it (or at least can be queried appropriately), so it's not just a blanket-ban on objects owned by SYS.

It occurs to me that you could manually do the deed using dba_objects, with a simple select statement such as: select object_name, object_id from dba_objects where object_id=61; In my case, this returned a segment name of SEQ$ -and the dollar sign is enough there to tell me it's all data dictionary magic, and not something I need to worry my head about (though knowing whether I should be converting back and forth between Hex and decimal is beyond me right now).

One final thought: Log Miner will throw a wobbly if it encounters Index-Organised tables, Clusters, Refs, Varrays, nested tables or chained rows. If you've got transactions involving any of that little lot in your Redo Stream, I wouldn't expect it to be analysed very neatly.

Hope that helps a little
HJR
>
> Sent via Deja.com http://www.deja.com/
> Before you buy.
Received on Sat Oct 28 2000 - 23:36:05 CDT