Re: Archive logs

"Ethan Post" <Blah_at_Blah.com> wrote in message news:BzBY6.60986$%a.3064329_at_news1.rdc1.sdca.home.com...
> From the way the question is phrased I think Joel could possibly think
 that
> a checkpoint only occurs when there is a log file switch. It is possible
> that checkpoints are already occuring more frequently than that. Log
> switches once per hour not a bad rate to be switching at. You could lose
> the ability to revcover if lose all of your current redo log files, which
> should be very unlikley if you are mirroring them. Check out
> LOG_CHECKPOINT_INTERVAL and following the links in the "See Also" section
> for other parameters. This will allow you to increase your checkpoint
> interval which will decrease instance recovery time if you have an
 instance
> failure.

Ethan, I may be getting confused... but are you saying here that *increasing* the intervals between checkpoints will *decrease* the time needed to perform Instance recovery? Because if so, regrettably, the precise opposite is actually the case.

Instance recovery means SMON replays all transactions in the current redo log hanging off the point of the last checkpoint. So if you checkpoint every 3 seconds, there will be no more than three seconds'-worth of transactions to replay. But if you checkpoint every 3 hours, then (absent a log switch) there will be 3 hours of transactions to replay...hence Instance recovery takes longer.

Now, I know that log_checkpoint_interval is measured in O/S blocks, not time, but the principle remains the same... the fewer the checkpoints, the longer Instance recovery (but the better general database performance).

I've probably just misunderstood your phrasing here, though...

Regards
HJR
>It will also ensure those changes stored in the current redo log
> files get out to the data files. However, I think your better off with
 less
> checkpointing and just ensuring you never lose your online redo log files.
>
> http://www.oradoc.com/ora817/server.817/a76961/ch189.htm#55753
>
> - Ethan Post
>
> "Daniel A. Morgan" <Daniel.Morgan_at_attws.com> wrote in message
> news:3B323887.4C787091_at_attws.com...
> > Joel wrote:
> >
> > > Our checkpoints/logfiles are changing about every hour during peak
 times. Any
> > > way we can recover that hour if something blows up?
> >
> > Depending on what blows up ... No.
> >
> > Change your checkpoint interview.
> >
> > Daniel A. Morgan
> >
>
>
Received on Mon Jun 25 2001 - 07:07:49 CDT