What about Doc ID 435887.1?
Charlie B, August 17, 2012 - 9:39 am UTC
Tom -
Thanks for looking into this.
We've always forced log rolls every hour, as part of our disaster recovery planning. A cron script kicks off the SQL statement for all our production db's.
Starting with 10.2, if my memory serves me right, we started getting these messages in most of our prod db's. We'd never seen them before the upgrade except during times of high db activity. Now they usually, but don't always, seem to occur in db's with low levels of activity.
ML doc id 435887.1 tells us that "this is expected behavior according to development." That's a shame, because we used to be able to ID db's with log performance issues in part due to the messages in the alert log. Now it's much harder: these "false positives" generate a lot of noise in the alert log.
Is there something we can do on the production side to prevent forced log rolls from generating these messages?
I learned something new - thanks!
Charlie B, August 20, 2012 - 12:44 pm UTC
A reader, October 17, 2014 - 10:08 am UTC
The note 435887.1 explains everything, so I think Tom's answer is not accurate.
From my understanding, This has nothing to do with performance, nothing to do with dirty data buffer not flushed to disk. It's about part of log buffer or private strand not flushed to the current redo log.