Gerald, in the absence of Oracle Data Guard or other high availability
options that feed off the redo log stream you generally want the total
number of archived redo logs generated per day to be somewhere between
24 to 48 similar to what Frank suggested. If you ever have to do a
fully manual recovery where you have to feed in each and every log name
you will know one reason why. But mostly you do not want multiple
checkpoints to overlap, that is, you really do not want a new log
switch and checkpoint to start if the prior checkpoint has not
finished. Nor do you want to fill all the online redo logs to fill
before the oldest log has been achieved to disk. Both of these
problems result in warning error messages being written to the alert
log. If your alter log is clean you are probably Ok with the rapid
switches if the total number of archived redo logs being generated is
not too high and if the number of times per day where you observe this
condition in the alert log is limited. That is you may need to size
the redo logs based on a peak processing time to prevent the problems
mentioned above even if it results in less total logs than would be
ideal.
HTH -- Mark D Powell --
Received on Thu Dec 29 2005 - 14:29:13 CST
