If you look in your alert file (on my system
%oracle_home%\rdbms73\trace\alrtorcl.ora), you will see messages giving the
times of each redo log switch. During a busy period the rule of thumb is
that these should be around every 30 minutes. There are three points to
guide your choice:-
- Log switches impose a small overhead, so you don't want them too often.
- If you are running in archivelog mode, the system needs enough time to
copy each log to the archive destination while it is not in use. If a log
has not been archived when the system wants to reuse it, the database will
stop dead:-<<<
- If your disk goes down and you want to recover using the archive logs,
the latest point you can get to is the time of the last log switch.
Therefore, the shorter the time between log switches the closer you can
get to recovering all changes to the database.
Received on Wed Aug 20 1997 - 00:00:00 CDT