Horton

Re: BUG #15460: Error while creating index or constraint

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

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Paul van der Linden <paul.vanderlinden@mapcreator.eu>
Cc: Peter Geoghegan <pg@bowt.ie>, PostgreSQL mailing lists <pgsql-bugs@lists.postgresql.org>, Heikki Linnakangas <hlinnaka@iki.fi>
Date: 2018-10-29T15:31:10Z
Lists: pgsql-bugs 
Paul van der Linden <paul.vanderlinden@mapcreator.eu> writes:
> Trace_sort is attached
> ...
> LOG:  performsort of 0 done: CPU: user: 250.03 s, system: 42.66 s, elapsed: 294.35 s
> LOG:  parallel external sort of 0 ended, 462886 disk blocks used: CPU: user: 250.03 s, system: 42.66 s, elapsed: 298.95 s
> LOG:  begin index sort: unique = f, workMem = 2096128, randomAccess = f
> LOG:  performsort of -1 starting: CPU: user: 0.00 s, system: 0.00 s, elapsed: 0.00 s
> LOG:  could not rmdir directory "base/pgsql_tmp/pgsql_tmp5088.0.sharedfileset": Directory not empty
> ERROR:  could not determine size of temporary file "0"

"performsort of -1"?  Seems a bit suspicious.  (And my complaint about
inadequate messaging holds here as well.  It's really unclear in this
trace which digit strings are to be read as numbers and which ones
are names of files/filesets.  At the very least, these messages violate
our message style guidelines: "When citing the name of an object, state
what kind of object it is.")

			regards, tom lane

Commits

  1. Have BufFileSize() ereport() on FileSize() failure.

  2. Use 64 bit type for BufFileSize().

  3. Adjust trace_sort log messages.

  4. Fix some sloppiness in the new BufFileSize() and BufFileAppend() functions.