Re: One archive log every minute

From: Igor Sereda <sereda_at_spb.runnet.ru >> Can someone give me a hint on how to reduce the speed of generating archived redo logs << Also mentioned that only a few jobs are on the system and implied that the jobs are read only or at least low update.

1. What version of Oracle are you on? There was a release of 7.2 that caused read data to be marked as dirty so it was had to be wrote even though it was unchanged
2. The dbms_alert package uses the dbms_lock package which issues commits.
3. What are the size of your redo logs?

Suggestions:

1. Increase the size of you redo logs to be large enough to hold some percentage of you total daily redo output such as 1/24 or 1/48 (which would give you a switch every half hour on average).
2. Check that the init.ora parameters log_checkpoint_interval should be set to a number of OS blocks that exceeds the size of a the redo log. Note - only many system this is 512 byte size blocks and not the official OS blocksize.

Also check that log checkpoint_timeout is either missing, set to 0, or has a value greater than the number of seconds that you want to average switching log, i.e., 30 minutes, an hour etc...

3) Scan the jobs being ran to make sure none of them issues an 'alter system checkpoint' or 'alter system switch logfle' command.

If these suggestions do not help then you may want to re-post with your Oracle version number, OS version, and redo log data.

Mark Powell -- Oracle 7 Certified DBA
- The only advice that counts is the advice that you follow so follow your own advice - Received on Sun May 24 1998 - 10:53:41 CDT