Re: Online backup: Backup online redologs?

On Tue, 12 Jun 2001, Howard J. Rogers wrote:

> > Couchman has been wrong many times, but here are his comments from the 8i
> > OCP book:
 

> > "Oracle uses SCNs in control files, datafile headers, and redo records.
> > Every redo log file has both a log sequence number and low and high SCN.
> > The low SCN records the lowest SCN recorded in the log file, while the
> > high SCN records the highest SCN in the log file. CKPT writes those
> > numbers to the datafiles and to the control file. The checkpoint number is
> > also written to the redo log file. When the database starts, all
> > checkpoint sequence numbers in all datafiles, redo log files, and control
> > files must match. If they do not, Oracle will not start, and you must
> > perform media recovery on your database to get the files synchronized and
> > in a consistent state." (p. 669)
 

> > This seems to imply to me that interrupting an in-flight checkpoint with a
> > shutdown abort can be deadly.
 

> No, it's not deadly.
 

> Shutdown abort requires an Instance Recovery every time, but that's not a
> problem, since the relevant redo will be available in the current redo log.
> There's nothing terribly magical about a shutdown abort: pull the plug out
> of your server and you've just performed a functionally-equivalent shutdown.
> So if shutdown abort could be 'deadly' then so could power failure be... and
> I doubt Oracle would have such a large market share if its principle product
> couldn't cope with intermittent power outages.
 

> I think the real issue is that there is a difference between the SCN
> assigned to each transaction as it's committed, and which gets written into
> the logs, and the sequence numbers that CKPT writes into the headers of all
> files at a checkpoint. Media failure means those CKPT numbers will be out
> of synch. Instance Failure means CKPT's numbers are perfectly consistent
> with each other... but there are still SCNs assigned *after* the last CKPT
> number.

Well, the warning above sounds pretty scary. I do get into situations where a shutdown immediate hangs (somtimes if I wait 5 or 10 minutes, it will complete). I suppose that I could select from v$process in such a situation and I might find something to kill, but I've never tried it.

In such a situation, with a hung shutdown immediate, I would like some method of ensuring that this information is synchronized between the datafiles before I proceed with an abort.

If Oracle doesn't want me to get skittish about this warning, then they shouldn't put such inflammitory things in the documentation. It is hard to accept such material as dogma when it is so full of holes.

p.s. Guess what, Howard? I passed the 8i OCP exam #2 last week. Heaven

     help us!

   / Charles J. Fisher                   | "Waste no more time arguing what  /
  /  cfisher_at_rhadmin.org                 |  a good man should be. Be one."  /
 /   http://rhadmin.org                  |   -Marcus Aurelius              /
---------------------------------------------------------------------------