David G. Johnston wrote:
> To be more clear, you need to run a query that will complete in our
> lifetime (and without an OOM error) with all three of ANALYZE, BUFFERS, and
> TIMING specified for the EXPLAIN.
I think the problem is pretty clear. The plan is sensible yet the
result doesn't seem to be. Why do you think using up all the memory is
a sensible result here?
Jaime Casanova suggested that maybe the @> operator have memory leaks.
Or perhaps the GIN index machinery that's using them.
--
Álvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services