On Mon, Aug 27, 2018 at 11:38 AM, Alexander Korotkov <a.korot...@postgrespro.ru> wrote: > The aspect I'm more concerned here about is whether we miss ability > for detecting some of IO errors, if we don't distinguish new pages > from pages whose tuples were removed by vacuum.
My main concern is correctness. If we ever have valid-looking buffers in shared_buffers after the corresponding data has been truncated away on disk, we've got to make sure that nobody ever confuses one of them with an actually-valid buffer. Reading over your algorithm, I can't convince myself that you have that case nailed down tightly enough. -- Robert Haas EnterpriseDB: http://www.enterprisedb.com The Enterprise PostgreSQL Company
