Horton

Re: hyrax vs. RelationBuildPartitionDesc

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

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Robert Haas <robertmhaas@gmail.com>
Cc: Alvaro Herrera <alvherre@2ndquadrant.com>, Amit Langote <Langote_Amit_f8@lab.ntt.co.jp>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Date: 2019-04-12T19:47:19Z
Lists: pgsql-hackers

Attachments

Robert Haas <robertmhaas@gmail.com> writes:
> As a parenthetical note, I observe that relcache.c seems to know
> almost nothing about rd_partcheck.  rd_partkey and rd_partdesc both
> have handling in RelationClearRelation(), but rd_partcheck does not,
> and I suspect that's wrong.  So the problems are probably not confined
> to the relcache-drop-time problem that you observed.

I concluded that that's not parenthetical but pretty relevant...

Attached is a revised version of Amit's patch at [1] that makes these
data structures be treated more similarly.  I also added some Asserts
and comment improvements to address the complaints I made upthread about
under-documentation of all this logic.

I also cleaned up the problem the code had with failing to distinguish
"partcheck list is NIL" from "partcheck list hasn't been computed yet".
For a relation with no such constraints, generate_partition_qual would do
the full pushups every time.  I'm not sure if the case actually occurs
commonly enough that that's a performance problem, but failure to account
for it made my added assertions fall over :-( and I thought fixing it
was better than weakening the assertions.

I haven't made back-patch versions yet.  I'd expect they could be
substantially the same, except the two new fields have to go at the
end of struct RelationData to avoid ABI breaks.

Oh: we might also need some change in RelationCacheInitializePhase3,
depending on the decision about [2].

			regards, tom lane

[1] https://www.postgresql.org/message-id/036852f2-ba7f-7a1f-21c6-00bc3515eda3@lab.ntt.co.jp
[2] https://www.postgresql.org/message-id/5706.1555093031@sss.pgh.pa.us

Commits

  1. Load relcache entries' partitioning data on-demand, not immediately.

  2. Prevent memory leaks associated with relcache rd_partcheck structures.

  3. Don't copy PartitionBoundInfo in set_relation_partition_info.

  4. Further reduce memory footprint of CLOBBER_CACHE_ALWAYS testing.

  5. Fix some oversights in commit 2455ab488.

  6. Defend against leaks into RelationBuildPartitionDesc.

  7. Allow ATTACH PARTITION with only ShareUpdateExclusiveLock.