Re: Oracle Doc Error

"Daniel Roy" <danielroy10junk_at_hotmail.com> wrote in message news:3722db.0401221803.5dbeb553_at_posting.google.com...
> Thank you Paul for your reply. I should first give a bit more details
> on my setup: I don't use archivelog (this is only my "toy" laptop
> database), and I don't multiplex my redo logs either. I don't fully
> understand your comments about the control returning to the user. It
> seems to me that if I issue "alter system checkpoint", the control
> will return after the checkpoint completed, and if I issue "alter
> system switch logfile", the control returns after the switch occurred.
> As far as I understand, a log switch is independant of a checkpoint
> (one doesn't imply the other). I therefore think that Oracle should
> say that a checkpoint (not a log switch) is necessary before a redo
> log drop can happen.
>

Hi Daniel

Yes and no.

If you simply issue a checkpoint but don't perform the log switch, then you of course still can't drop the group.

*Both* the log switch (thereby no longer making the log group "CURRENT") and it's full checkpoint (thereby no longer making the log group "ACTIVE") must occur before the redo log can be dropped. The completed checkpoint ensures that the redo log is no longer required in the event of an instance failure (which would make dropping it a somewhat dangerous thing).

Depending on how you've configured your checkpoints, you can either wait for the implicit full background checkpoint to occur that's generated when you perform the log switch when you don't use incremental checkpointing, wait for the checkpoint marker to pass the redo log in question when you use incremental checkpointing, or force the issue with an explicit checkpoint with the alter system command.

In summary, the redo log must be "INACTIVE" and archived (assuming you have the DB in archivelog mode) before it can be dropped.

Hope this makes sense.

Cheers

Richard Received on Sun Jan 25 2004 - 03:59:43 CST