Re: Error advancing redo logs

When you switch log files, Oracle checkpoints. That is to say it writes all the dirty buffers out to the datafiles, updates the headers in the datafiles, control files with the latest SCN. Oracle was not done with with process when it needed a new redo log file.

You can read about this in the Oracle Server Tuning Manual Tuning Checkpoints section.

This message you encountered means Oracle slowed down or paused a bit. If this happens very infrequently and performance is not a problem, then you can ignore it.

If it presents a performance problem, you can tune checkpoints. You can specify a CKPT process be started to handle checkpoints, make checkpoints happen more often than log switches. You can also specify a larger batch of dirty buffer blocks be written out every 3 seconds when DBWR wakes up to write out some dirty blocks. If a larger batch of dirty blocks is written out every 3 seconds, checkpoint at log switches will complete more quickly. Indeed if the database is not very active, eventually the 3 second batches will write all the dirty buffers out to the data files.

Alternately, you can tune your redo log files to avoid these situation.

Doug Cowles (dcowles_at_bigfoot.com) wrote:
: Can anyone give me any suggestions as to why the alert log would have
: several entries
: indicating "Thread 1 cannot allocate new log....checkpoint not complete"
: etc., ?
: Redo log files are on different disks, no known problems with them...
: any suggestions
: appreciated.

: dcowles_at_bigfoot.com

-- Received on Sat Oct 31 1998 - 15:18:55 CST