On Mon, Aug 12, 2019 at 3:07 PM Alvaro Herrera <alvhe...@2ndquadrant.com> wrote: > How likely is it that we would ever be able to release memory from a > Sort (or, say, a hashjoin hash table) when it's done being read, but > before completing the whole plan? As I understand, right now we hold > onto a lot of memory after such plans have been fully read, for no good > reason other than executor being unaware of this. This might not be > directly related to the problem at hand, since it's not just parallel > plans that are affected.
Being able to do that sort of thing was one of my goals in designing the ExecShutdownNode stuff. Unfortunately, it's clear from this bug report that it's still a few bricks short of a load, and Tom doesn't seem real optimistic about how easy it will be to buy those bricks at discount prices. But I hope we persist in trying to get there, because I don't like the idea of saying that we'll never be smart enough to know we're done with any part of the plan until we're definitely done with the whole thing. I think that's leaving too much money on the table. -- Robert Haas EnterpriseDB: http://www.enterprisedb.com The Enterprise PostgreSQL Company
