Re: Oracle/VMS Backup

From: Norm Soley - That's MISTER Unix geek to you <soley_at_trop32.enet.dec.com>
Date: Fri, 20 May 1994 03:19:03 GMT
Message-ID: <1994May20.031903.2997_at_peavax.mlo.dec.com>


In article <1994May19.013114.17205_at_nntpxfer.psi.com> dmausner_at_brauntech.com writes:
>
>In article <768907062snz_at_hiredata.demon.co.uk>, <Parker_at_hiredata.demon.co.uk>
>writes:
>> 2) Using BEGIN & END BACKUP:
>>
>> a) what does Oracle do in this mode?
>
>it avoids writing to the tablespace during backup mode. all changes are
>posted to the redo logs and the rollback segments during this time. when
>backup mode is removed, the database writer updates the tablespace. this
>explains why it's safe to backup a file in backup mode: no writing is
>performed.

I was under the impression that it was a little different. The DBWR continues to write to the file throughout the backup. BEGIN and END BACKUP change the way that log checkpoints are recorded in the control files (or is it the control block of the datafile??? that makes more sense) so that, although checkpoints are happening they are not recorded. This avoids the running out of re-do log space problem. In the event of a failure when you attempt to remount the recovered file Oracle will detect that it is way out of sync and attempt to roll the file forward from the archive logs or at least inform you that manual recovery is necessary. Since the redo-logs contain after image data the file need not be in a consistant state when recovered from backup since, once all the logs are applied only the most recent change remains. If the change was already written it get's written again, multiple changes...well you get the idea.

-- 
Norm Soley, Leash trained UNIX geek, Digital Consulting/Digital Equipment Canada
soley_at_trooa.enet.dec.com  
			  | Opinions expressed are mine and do not reflect those
                          | of Digital Equipment Corporation or my cat Marge. 
Received on Fri May 20 1994 - 05:19:03 CEST

Original text of this message