Re: Cant bring up cold backup

  • From: Mark Bole <makbo@xxxxxxxxxxx>
  • To: mwf@xxxxxxxx
  • Date: Mon, 06 Jun 2005 07:02:44 -0700

There is a lot to this topic, and it comes up with some regularity in a 
number of forums. Here are two comments up front, with a few additional 
comments interleaved in your message as well.  By no means does this 
cover it all.

1) RMAN is preferable to user-managed backup and recovery from version 
9i onward, making these types of errors less likely to happen.

2) Terminology is important: a "closed, consistent, whole database 
backup" is a more useful description than "cold backup".  Restoration 
and recovery are two different things. And so on.

-Mark Bole


Mark W. Farnham wrote:
[...]
> How exactly is using a created or backup controlfile preventing complete
> recovery?

I'll back away from that claim, based on a document I just re-read 
(Howard J. Rogers "Backup and Recovery" (c) 2000), which states that the 
Oracle documentation is in error on this point.

The important point I wanted to make was, a current control file has SCN 
information in it, while a created controlfile (at time of creation) 
obviously does not.

Update: I just performed a test, using only the datafiles of a cleanly 
shut down instance.  I created a controlfile beginning with:

SQL> CREATE CONTROLFILE set database binc99 RESETLOGS  NOARCHIVELOG ...
Control file created.
SQK> alter database open resetlogs;
Database altered.

and found the following in the alert log.  So this shows that no media 
or instance recovery should be required even in this case, despite the 
warning about "incomplete recovery".  The OP's problem most likely is 
somewhere else.

Mon Jun  6 06:46:48 2005
Completed: CREATE CONTROLFILE set database binc99 RESETLOGS
Mon Jun  6 06:46:55 2005
alter database open resetlogs
RESETLOGS after incomplete recovery UNTIL CHANGE 2277844

Note the the SCN number must be coming directly from the tablespace 
headers, since it obviously was not in the controlfile or redo logs.

> And since this is
> recovery, not backup, you've brought in the online logs from the former
> location before that former database was restarted, right?

Oracle best practice is to never backup your online redo logs (multiplex 
yes, backup no).

http://download-west.oracle.com/docs/cd/B10501_01/server.920/a96519/strategy.htm#1004884

But yes, you can do it, and in this scenario, if everything was copied 
(current controlfiles, online redo logs, etc), then the copy on the new 
server could indeed be started and opened with no special handling.  See 
my other response for a recommended way for the OP to proceed which 
observes this best practice.

> Although the whole create tracefile thing was a bell or
> whistle added to avoid folks having to get the rename datafile syntax right
> and in the exactly right sequence in the recovery startup

It serves quite a few more purposes than that!

> Then again, I have little or no experience
> with recovering noarchivelog databases.

I don't do it much these days either, nor do many others, I suspect. 
Whatever business need there is (the OP did not state), there are 
probably better ways to copy, clone, replicate [your term here] a database.

>>
>>  I am trying to bring up a database from a cold backup, Using the same
>>dbname
>>
[...]
>>3 - use a alter database backup control file to trace, edit to point to
> the
>>new datafiles


--
//www.freelists.org/webpage/oracle-l

Other related posts: