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 questions

Re: Hot backup questions

From: Doug Cowles <dcowles_at_nospambigfoot.com>
Date: Wed, 27 Oct 1999 12:25:37 -0400
Message-ID: <38172781.78400DDC@nospambigfoot.com>


Thanks for the thread everyone. It's all coming back to me now. I do remember this
piece of information about hot backups but I had forgotten it. The problem was two-fold. A hot-backup was having trouble writting to tape, and abended without taking a tablespace out of backup mode. In addition, a sysadmin when rebooting the box didn't shut down Oracle cleanly (I think), so the frozen SCN number required an application of some archive logs to fix itself, even though it wasn't really out of synch as these posts suggest, but more a few SCN numbers behind. I'm not sure what happens if you bring
a database down cleanly while in hot backup mode, but I'll try it. Maybe Oracle was shut
down cleanly after all.
Thanks again,
Dc.

Brian Peasland wrote:

> I did some checking on this. I found it odd that the Oracle Unleashed
> (by SAMS Publishing) and the Oracle Backup & Recovery (Oracle Press)
> books both state that datafiles are not written to during hot backups.
> These were the sources of my previous comments. After Jeremiah's post, I
> read the article alluded to in his URL. I found it to be a well written
> article. Thanks Jeremiah! But there is always a danger of taking things
> from the web and citing them as truth. No offense intended here. So I
> did some more digging. In the Oracle Backup and Recovery documentation
> provided with Oracle 8.0.5, I found the following:
>
> "A backup of an open database is inconsistent because portions of the
> database (hence, datafiles in the database) are being modified and
> written to disk while the backup is progressing. "
>
> Thanks for clearing up my (and hopefully others) misconceptions about
> this.
>
> Thanks!!!!
> Brian
>
> Jeremiah Wilton wrote:
> >
> > Brian Peasland wrote:
> > >
> > > Jeremiah Wilton wrote:
> > > >
> > > > Tablespaces in backup mode are not out of sync, so they have no need to
> > > > resync when they are taken out of backup mode. Blocks are written to
> > > > the datafiles during backup mode as they are during normal operation.
> > >
> > > Is this true? Blocks are written to the datafiles during backup mode? If
> > > so, then the datafile is changed while it is being backed up resulting
> > > in an invalid backup of the datafile.
> >
> > The backup datafile copies are inconsistent (not invalid), but then you
> > apply logs to them during recovery to make them consistent.
> >
> > > If memory serves me right, while a
> > > tablespace is in backup mode, any changes to it are written to the redo
> > > logs.
> >
> > Changes written to tablespaces not in backup mode are also written to
> > the redologs. Redologs record all changes.
> >
> > > And the changes are written a block at a time. This means that the
> > > redo logs might fill up faster if there is high volume.
> >
> > The logging of full blocks instead of just changed data for tablespaces
> > in backup mode allows any split blocks (blocks read by the backup at the
> > same time they were being written by dbw0) to be resolved during
> > recovery by completely replacing the changed blocks in the datafiles
> > with the good block images from the redologs.
> >
> > > After the
> > > tablespace is taken out of backup mode, the changes are then applied to
> > > the datafiles.
> >
> > That part doesn't happen. How come the end backup command returns
> > instantly, even if the tablespace has been backing up for several hours?
> > Over the years, so many people have not believed me, and asked for
> > "proof" that datafiles are written during backup mode that I have
> > prepared a repeatable proof that shows that Oracle writes data to
> > datafiles in backup mode. It can be found at the following URL:
> >
> > http://www.wolfenet.com/~jeremiah/hot-backup.html
> >
> > --
> > Jeremiah


Received on Wed Oct 27 1999 - 11:25:37 CDT

Original text of this message

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