Horton

Re: Out of Memory errors are frustrating as heck!

Alvaro Herrera <alvherre@2ndquadrant.com>

View thread View on postgr.es
From: Alvaro Herrera <alvherre@2ndquadrant.com>
To: Gunther <raj@gusw.net>
Cc: Jeff Janes <jeff.janes@gmail.com>, Tomas Vondra <tomas.vondra@2ndquadrant.com>, pgsql-performance@lists.postgresql.org, Tom Lane <tgl@sss.pgh.pa.us>, Justin Pryzby <pryzby@telsasoft.com>
Date: 2019-04-15T17:38:48Z
Lists: pgsql-performance

Attachments

On 2019-Apr-15, Gunther wrote:

> #0  AllocSetAlloc (context=0x1168230, size=385) at aset.c:715
> #1  0x000000000084e6cd in palloc (size=385) at mcxt.c:938
> #2  0x000000000061019c in ExecHashJoinGetSavedTuple (file=file@entry=0x8bbc528, hashvalue=hashvalue@entry=0x7fff2e4ca76c,
>     tupleSlot=0x10856b8, hjstate=0x11688e0) at nodeHashjoin.c:1277
> #3  0x0000000000610c83 in ExecHashJoinNewBatch (hjstate=0x11688e0) at nodeHashjoin.c:1042

Seems that ExecHashJoinGetSavedTuple stores a minimalTuple and sets the
shouldFree flag to "true", and then in ExecHashJoinNewBatch, callee
ExecFetchSlotMinimalTuple sets shouldFree to false inconditionally when
the slot uses minimal tuple ops.  Maybe that's correct, but it does
sound like a memory leak is not entirely impossible.  I wonder if this
fixes it, without causing crashes elsewhere.

-- 
Álvaro Herrera                https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

Commits

  1. Consider BufFiles when adjusting hashjoin parameters

  2. Allocate hash join files in a separate memory context