Re: switch logfile, checkpoint, active and inactive logfile

Wow !!! Nice questions. Let me try to answer these. I may not be entirely correct so, anyone, feel free to correct me (just don't beat me up ;) )

1. switch logfile DOES trigger the checkpoint instaneously, but the checkpoint may not be finished. So now when the headers of control files and datafiles are updated? after DBWn have done job or checkpoint is initiated?
2. If LGWR jumps to the next logfile (now becomes current) without waiting for Checkpoint to be done, the checkpoint for instance recovery is in the current logfile, or previous, but still active logfile?

I will take 1 & 2 together. Switching a logfile essential takes out the current logfile from being current to some other state (inactive,active, etc) and brings in another available redolog file to current state. The baggage that comes with it, which is checkpoint works sort of on its own. It has its own starting time and its own ending time and the way the headers are updated depends on the checkpointing process. So till the point it is complete the last checkpoint information is in one of the "active" redologs. The rationale being, if the checkpoint information is in "current" redologs, other redos are no longer required for instance recovery hence should be marked as "inactive".

3) Instance recovery roll forward to the latest checkpoint, so when this point of check point is marked? in the begining of checkpoint or the end?

End of the checkpoint

//tc

"Niy" <niy38_at_hotmail.com> wrote in message news:55dd405a.0304241131.7832ca24_at_posting.google.com...
> >
> > returns control to you after the logfile is switched, which does not
> > necessarily mean that it will complete the checkpoint by then. It is
 quite
> > possible that the checkpoint completes some time after the switch or
 perhaps
> > does not complete at all.
>
> According to Oracle8i Concepts:
>
> Checkpoint Process (CKPT)
>
> When a checkpoint occurs, Oracle must update the headers of all
> datafiles to record the details of the checkpoint. This is done by the
> CKPT process. The CKPT process does not write blocks to disk; DBWn
> always performs that work.
>
>
>
> Now I got other questions:
> 1) switch logfile DOES trigger the checkpoint instaneously, but the
> checkpoint
> may not be finished. So now when the headers of control files and
> datafiles
> are updated? after DBWn have done job or checkpoint is initiated?
>
> 2) If LGWR jumps to the next logfile (now becomes current) without
> waiting for
> Checkpoint to be done, the checkpoint for instance recovery is in
> the
> current logfile, or previous, but still active logfile?
>
> 3) Instance recovery roll forward to the latest checkpoint, so when
> this point
> of check point is marked? in the begining of checkpoint or the end?
>
>
> Thanks for the reply. Until this point (not checkpoint) I feel the
> difficulty
> of Oracle.
Received on Thu Apr 24 2003 - 22:58:05 CEST