Re: Out of Memory errors are frustrating as heck!
Jeff Janes <jeff.janes@gmail.com>
From: Jeff Janes <jeff.janes@gmail.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Gunther <raj@gusw.net>, Justin Pryzby <pryzby@telsasoft.com>, Tomas Vondra <tomas.vondra@2ndquadrant.com>,
Alvaro Herrera <alvherre@2ndquadrant.com>, pgsql-performance@lists.postgresql.org
Date: 2019-04-22T17:15:10Z
Lists: pgsql-performance
On Sat, Apr 20, 2019 at 4:48 PM Tom Lane <tgl@sss.pgh.pa.us> wrote: > Gunther <raj@gusw.net> writes: > > and checked my log file and there was nothing before the call > > MemoryContextStats(TopPortalContext) so I don't understand where this > > printf stuff is ending up. > > It's going to stdout, which is likely block-buffered whereas stderr > is line-buffered, so data from the latter will show up in your log > file much sooner. You might consider adding something to startup > to switch stdout to line buffering. > Is there a reason to not just elog the HJDEBUG stuff? With some of the other DEBUG defines, we will probably be using them before the logging system is set up, but surely we won't be doing Hash Joins that early? I think we could just get rid of the conditional compilation and elog this at DEBUG1 or DEBUG2. Or keep the conditional compilation and elog it at LOG. Cheers, Jeff
Commits
-
Consider BufFiles when adjusting hashjoin parameters
- a1b4f289beec 18.0 landed
-
Allocate hash join files in a separate memory context
- 8c4040edf456 16.0 landed