> From: Greg Smith [mailto:gsmith@xxxxxxxxxxxxx] > Sent: Monday, May 19, 2008 9:03 PM > > So, yes, in 8.3 it's possible that you can have sequential > scans of large > tables or the VACUUM data pass through the buffer cache, but > not remain in > it afterwards. I didn't think George would ever run into this in the > specific example he asked about because of (1). This > behavior only kicks > in if you're scanning a table large relative to the total > shared buffer > cache and that didn't seem like an issue in his case. Correct -- the tables in this example were tiny, shared buffers are large, and, in any case, I am still on 8.1... George
