I thought it might be helpful to show what sar -d returns during a disk issue. Sorry, I only have a screenshot I took during the problem. Click on the smaller image to see the full size jpeg.
This was a “sar -d 5 2” – two five second intervals. Here is what this looks like now that the disk subsystem is functioning again:
When I first checked on the problem database the sessions were waiting on log file sync. I did an insert into a table and the insert ran quickly, but the commit took like 27 seconds. I believe this is because the block I inserted into wasn’t immediately written to disk but the commit forces a write to the redo log. So, anything doing a commit was taking forever. One thing I’m confused about is why the long time is avwait while the avserv is short. Maybe this is because the current request is taking forever and isn’t included in avserv whereas the wait time for the next request is included in avwait, but that is just a guess. Maybe if I did a longer interval for sar -d then the avserv would have been higher, but I’ve only got the screenshot I took at the time.