Re: IndexJoin memory problem using spgist and boxes

Tom Lane <tgl@sss.pgh.pa.us>

From: Tom Lane <tgl@sss.pgh.pa.us>
To: Anton Dignös <dignoes@inf.unibz.it>
Cc: Alexander Kuzmenkov <a.kuzmenkov@postgrespro.ru>, pgsql-hackers@lists.postgresql.org
Date: 2018-03-04T19:46:22Z
Lists: pgsql-hackers
=?UTF-8?Q?Anton_Dign=C3=B6s?= <dignoes@inf.unibz.it> writes:
>> Looking at the patch, I see that you changed the lifetime of the temporary
>> context from per-tuple to per-index-scan. It is not obvious that this change
>> is correct.

> The problem before this patch was that the traversalMemoryContext in
> this function was set to per-query lifetime.
> The memory allocations in the per-query lifetime caused this high
> memory consumption.

Yeah ...

> I changed the temporary context to per-index-scan so that it can also
> be used for traversalMemoryContext.

But we have also had many complaints about leakage across a single index
scan, if it traverses many index entries.  I think you're just moving the
pain out of one use-case and into another.

			regards, tom lane


Commits

  1. Prevent query-lifespan memory leakage of SP-GiST traversal values.