Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
![]() |
![]() |
Home -> Community -> Usenet -> c.d.o.server -> Re: Hot and Cold backups
ora_lrnr_at_yahoo.com wrote:
> > > Hi,
> > >
> > > Here what is your archive mode? Is it noarchivelog? If it is noarchive
> > > then you need to restore all the cold backup and that's all ...so you
> > > lose one day worth of data.
> > >
> > > Now if it is archivelog then you need to restore and then recover
> > > database by applying the archivelogs so you will NOT lose any data, if
> > > you have preserved all your archive logs....
> > >
> >
> >
> > You will loose whatever transactions are in the (now dead/broken/lost)
> > active redo logs.
> > Having a backup of the online redo logs will not help as they contain
> > transactions from yesterday.
> > Just open resetlogs. You cannot recover past the last archivelog, or the
> > last unarchived online log (not current) that may still be available on your
> > disk.
> >
>
>
> one thing that has always confused me is the ALTER DATABASE CLEAR
> LOGFILE command. If it is for clearing corrupt logfiles, you will not
> be able to archive those log files. Wouldnt Oracle crash? Wouldnt you
> require a partial recovery followed by resetlogs after a logfile goes
> corrupt. Why do you need to have the command CLEAR LOGIFLE? If it is
> corrupt isnt the database screwed.
See the fine manual:
http://download-west.oracle.com/docs/cd/B10501_01/server.920/a96521/onlineredo.htm#1260
You could then keep going, and just do another online backup. The alert log will say which backups you wouldn't be able to recover completely from.
>
> Can someone explain when to clear an archived and/or unarchived logs vs
> do a partial recovery with open resetlogs?
Just because a redo log is corrupt doesn't mean your data is corrupt.
It just means you can't use it to recover (when you have to "re-do" the
transaction). And think on some ramifications here:
http://www.oracledba.co.uk/tips/lgwr_dilemma.htm So in general,
everything will be okay unless you use some 10G async commit thingy.
(also see
http://asktom.oracle.com/pls/ask/f?p=4950:8:::::F4950_P8_DISPLAYID:1415454871121#20061201507022
about what PL/SQL does). You may lose some committed data if you lose
some redo log _and_ have to recover. Oracle can be amazing about these
things sometimes, especially about figuring out which redo log to
archive when you have Oracle mirroring and corruption (as opposed to
hardware mirroring mirroring the corruption).
So you have to do a partial recovery and open resetlogs if you lose something beyond just a corrupt redo log.
jg
-- @home.com is bogus. Newtons book not fit to be printed: http://www.flickr.com/photos/edyson/668917/Received on Wed Jul 13 2005 - 17:30:15 CDT
![]() |
![]() |