Re: Hotbackup under VMS

From: Howard J. Rogers <howardjr_at_www.com>
Date: Mon, 12 Feb 2001 08:17:20 +1100
Message-ID: <3a870162$1@news.iprimus.com.au>

"Peter J. Holzer" <hjp-usenet_at_hjp.at> wrote in message news:slrn98do7h.pns.hjp-usenet_at_teal.h.hjp.at...
> On 2001-02-11 13:23, WCK <abramswee_at_yahoo.com> wrote:
> >just do on the svrmgrl first and do this;
> >
> >svrmgrl> alter tablespace xxxxx begin backup;
> >svrmgrl> ! cp /oradata/prod/proddatafile1.dbf /dev/rmt0 (or whichever u

 do
        ^^^^^ This doesn't look like VMS to me.


> >in vms to do file backup);

> >svrmgrl> alter tablespace xxxxx end backup;

> >
> >do this for the rest and don't forget the controlfiles.
>
> But more importantly:
>
> If I backup one data file after the other, How can I ensure that the
> backup is consistent?
>

It won't be consistent. It's guaranteed NOT to be consistent. Even internally within a single datafile, it won't be consistent.

But who cares? Because applying redo MAKES things consistent.

Provided you have ALL the redo (archived and online) since the time that you START your backup cycle, you can guarantee to be able to make consistent any part of your backup set, however long it takes to generate. And the business of putting tablespaces into hot backup mode before you start copying? Since that locks the timestamp in the header of each datafile within the tablespace, we know what the 'time' of the database was BEFORE you started copying files -and that means it's guaranteed to be able to make the internal mess which is your copied datafile consistent by the application of redo since that time.