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: Reading a comitted data before DBW writes it to the disk.

Re: Reading a comitted data before DBW writes it to the disk.

From: Michel Cadot <micadot{at}altern{dot}org>
Date: Fri, 2 Apr 2004 15:22:51 +0200
Message-ID: <406d69b9$0$12801$626a14ce@news.free.fr>

"Somu" <soumya_ongc_at_yahoo.com> a écrit dans le message de news:9666da71.0404020103.7e174d21_at_posting.google.com...
> Hi,
> I have some doubts on the above mentioned topic.
>
> As I understand, Oracle use Data buffer cache to store the data frm
> data file in memory for efficient transaction. Consider a case, when a
> row of a table is modified by a user,that means earlier row values are
> in roll back segment,current changed value in data buffer cache and
> info regarding this transaction in redo log buffer.Now suppose the
> user has committed the transaction.So redo log buffer content gets
> copied to redo log file.But as DBW writes to the data files at some
> time interval(viz 3 sec) and commit itself doesnot invoke any
> checkpoint so its clear that at this instance data files are not
> updated.Now suppose the user issues Select query to see that
> particular row fields ,now as par with data consistency Oracle will
> show the comitted data.
> Now my question comes into picture.Where from Oracle
> show this updated data? Apparently it seems from Data Buffer cache.So
> does this cache has information that explicitly says following info
> a)this block is changed but not comitted
> b)this block is changed and comitted but not written back
> to data files
> c)this block is changed and comitted and written back to
> data files
>
> Waiting for some positive response.
> Thanks and Regards
> somu

Have a look at Metalink note 40689 "ORA-01555 "Snapshot too old" - Detailed Explanation", url: http://metalink.oracle.com/metalink/plsql/ml2_documents.showDocument?p_database_id=NOT&p_id=40689.1

There is a detailed description of how "commited" flag is propagated.

Regards
Michel Cadot Received on Fri Apr 02 2004 - 07:22:51 CST

Original text of this message

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