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: Alvaro Herrera <alvherre@2ndquadrant.com>
Cc: Robert Haas <robertmhaas@gmail.com>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Date: 2019-03-13T18:36:13Z
Lists: pgsql-hackers 
Alvaro Herrera <alvherre@2ndquadrant.com> writes:
> You do?  Unless I misunderstood, your options are:

> 1. (the patch you attached) create a temporary memory context that is
> used for everything, then at the end copy the good stuff to CacheMemCxt
> (or a sub-context thereof).  This still needs to copy.

> 2. create a temp memory context, do everything there, do retail freeing
> of everything we don't want, reparenting the context to CacheMemCxt.
> Hope we didn't forget to pfree anything.

> How is any of those superior to what I propose?

I doubt that what you're suggesting is terribly workable.  It's not
just RelationBuildPartitionDesc that's at issue.  Part of what will
be the long-lived data structure is made by partition_bounds_create,
and that invokes quite a boatload of code that both makes the final
data structure and leaks a lot of intermediate junk.  Having to be
very careful about permanent vs temporary data throughout all of that
seems like a recipe for bugs.

The existing code in RelationBuildPartitionDesc is already pretty good
about avoiding copying of data other than the output of
partition_bounds_create.  In fact, I think it's already close to what
you're suggesting other than that point.  So I think --- particularly
given that we need something we can back-patch into v11 --- that we
shouldn't try to do anything much more complicated than what Robert is
suggesting.

			regards, tom lane

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.