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: Hot Backup Question (Again!)

Re: Hot Backup Question (Again!)

From: Thomas Kyte <tkyte_at_us.oracle.com>
Date: Tue, 23 Nov 1999 19:07:21 -0500
Message-ID: <ppam3schiao33eeidf1itnj4siudjjkfbh@4ax.com>


A copy of this was sent to MohammedB <mohammedbNOmoSPAM_at_iadb.org.invalid> (if that email address didn't require changing) On Tue, 23 Nov 1999 12:22:15 -0800, you wrote:

>Folks,
>
>Forgive me for asking again, I know someone recently asked this
>question....but (I wasn't paying complete attention)...
>
>As I understand it, when a tablespace is in hot backup mode, the SCN
>for the datafile headers are not updated, entire data blocks are
>written to the redo logs as opposed to just the redo records, and when
>there is a commit, the lgwr writes the contents of the log buffer to
>the online redo logs and eventually, the data blocks of the datafiles
>are updated by dbwr, just as when a tablespace is not in hot backup
>mode. My point being, even with on-line hot backups, changed database
>blocks do get written to the datafiles, right?
>

mostly -- entire blocks are written the first time they are modified in backup mode, the next time its back to redo vectors.

>So, having said that, I've read the following from a study guide:
>
>'When a tablespace is in backup mode, the log sequence numbering is
>temporarily halted and no more changes are made to it. Instead,
>transaction information is written to the redo log files. Once the
>tablespace is no longer in backup mode, all the information from the
>redo logs (including log sequence numbers) is applied to make it
>current.'
>

when a tablespace is in backup mode, the files associated with that tablespace don't have their headers updated. transaction information is written to redo (as is always the case) and once the tablespace is out of backup mode, the data file headers are updated. The datafiles are written to (checkpointed, flushed) continously throughout the backup.

>and also talked with someone who said the this:
>
>'While a tablespace is in backup mode, it might as well be offline. Any
>select statements for new data do come out of the datafiles, but
>anything changed or updated, goes to the redo log files.

wrong. You do not need archive redo log files to catch up an instance after its taken out of backup mode. If you had say 5-1meg log files and generated 10 meg of log during a backup -- it wouldn't work ('catching' up a datafile). Besides that, alter tablespace end backup is instantaneous, it doesn't do very much work. the datafiles are written to all the time.

>And of course
>for read consistency during concurrent transactions, the rollback
>segments keep the before image of any data. If there is data to be
>committed, it is committed to the redo logs. Then once the tablespace
>switches back to normal mode, the redo logs are applied. Basically, for
>new writes and changes, the redo logs are acting like temporary
>datafiles until recovery is done.'
>
>None of which I agree with because it does not make complete sense.
>

they are very wrong.

>Ok, so what am I missing?
>
>Any enlightenment appreciated.
>
>mkb
>
>
>
>* Sent from RemarQ http://www.remarq.com The Internet's Discussion Network *
>The fastest and easiest way to search and participate in Usenet - Free!

--
See http://osi.oracle.com/~tkyte/ for my columns 'Digging-in to Oracle8i'... Current article is "Part I of V, Autonomous Transactions" updated June 21'st  

Thomas Kyte                   tkyte_at_us.oracle.com
Oracle Service Industries     Reston, VA   USA

Opinions are mine and do not necessarily reflect those of Oracle Corporation Received on Tue Nov 23 1999 - 18:07:21 CST

Original text of this message

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