On Thu, Jan 17, 2019 at 1:57 AM Masahiko Sawada <sawada.m...@gmail.com> wrote: > The reason why I processed the tuples that became dead after the first > heap pass is that I was not sure the reason why we ignore such tuples > in the second heap pass despite of there already have been the code > doing so which has been used for a long time. I thought we can do that > in the same manner even in DISABLE_INDEX_CLEANUP case. Also, since I > thought that lazy_vacuum_page() is the best place to set them as DEAD > I modified it (In the previous patch I introduced another function > setting them as DEAD aside from lazy_vacuum_page(). But since these > were almost same I merged them).
The race you're concerned about is extremely narrow. We HOT-prune the page, and then immediately afterward -- probably a few milliseconds later -- we loop over the tuples still on the page and check the status of each one. The only time we get a different answer is when a transaction aborts in those few milliseconds. We don't worry about handling those because it's a very rare condition. -- Robert Haas EnterpriseDB: http://www.enterprisedb.com The Enterprise PostgreSQL Company
