Re: RMAN

On Aug 27, 2:40 pm, artme..._at_yahoo.com wrote:
> On Aug 27, 2:24 pm, sybra..._at_hccnet.nl wrote:
>
>
>
> > On Wed, 27 Aug 2008 12:01:19 -0700 (PDT), artme..._at_yahoo.com wrote:
> > >Sorry for the double post, but as we are reading and seeing BACKUPSET
> > >and BACKUPPIECE........we can see that a piece is part of a set. But,
> > >if you need to break up the set into pieces, why not just make smaller
> > >size sets??
>
> > A datafile can not span multiple backupsets.
> > A datafile can span multiple backuppieces.
> > You could configure 2G backuppieces (max filesize for 32-bit files on
> > some Unixes) and have a backupset of unlimited size.
>
> > >Also, how does this relate to FILESPERSET?
>
> > All files in a set will be
> > a) simultaneously opened
> > b) simultaneously backupped
> > c) get multiplexed in one backupset, so 1 block for datafile 1 is
> > followed by one block for datafile 2 etc.
>
> > If you have all your datafiles on few disks, NOT configuring
> > filesperset (Oracle by default tries to back up as much as possible)
> > create an I/O bottleneck.
> > You could argue filesperset always needs to be 1.
>
> > >We're all new to this.....
>
> > >Thank you
>
> > Hth
> > --
> > Sybrand Bakker
> > Senior Oracle DBA
>
> Thank you Sybrand,
>
> I'm sure this has a lot to do with performance. So, we'll take this
> and try to play and see some stats.

Sybrand,

Forgive me for this last question.

I think I understand backupset & backuppiece. Usually each backupset is allocated to a different channel.

Filesperset I'll have to do a bit more reading on as to understand why 3 files writing to 3 different backup sets vs 3 files writing to 1 backup set makes a difference.

MAXSETSIZE sets the maximum size per backupset, I assume. Would setting both filesperset & maxsizeset conflict? I'm trying to think about this. If you have 12 files, and you say you want 4 files per set, thus 3 sets, and a maxsize of 50G per set.......you may get more than 3 sets, defined by the previous filesperset value......

I'm probably wrong..... Received on Wed Aug 27 2008 - 15:13:46 CDT