"posting the previous set of waiters" as You said. However I suppose also, that if it would be so, we would not experience jumps of "log file sync" waits while switching between logs, but rather almost continuous increase of their value ...well, I thought you had an abnormally high LFS wait all the taime and spikes under switches. That would be normal on processor-staturated systems (that are also logging-intensive). -- //www.freelists.org/webpage/oracle-l