A few years ago I wrote a series of posts about various scenarios of log file sync (LFS) waits. A few days ago, I have had an issue which added yet another scenario to that already diverse collection, and I thought it would be potentially of interest for others in the Oracle community. To set the scene, it’s the same 18c (18.10 to be more accurate) 3-node X6-2 Exadata cluster where most of issues I describe in my blog occur. At a certain moment, I noticed LFS go up quite a bit, with log file parallel write (LFPW) also showing signs of increase, but a much more moderate one. I’ve got a script to graph wait event stats for multiple wait events (e.g. using a regular expression pattern for their names), below you can see a panel summarising average wait times for events containing “log file” or “redo” in their names.

Continue reading “Log file sync and and LGWR redo wait for copy wait events” →