Re: test_fsync file overrun

From: Bruce Momjian <bruce(at)momjian(dot)us>
To: Jeff Janes <jeff(dot)janes(at)gmail(dot)com>
Cc: pgsql-hackers <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: test_fsync file overrun
Date: 2009-09-21 20:21:11
Message-ID: 200909212021.n8LKLB504952@momjian.us
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Jeff Janes wrote:
> test_fsync in tools/fsync pre-creates a 16MB file. If it is given a number
> of iterations greater than 1024 (like one might use if trying to see what
> happens when NVRAM gets filled, or on a journaling file system), than one of
> the writes being timed will have to extend the size of the pre-created test
> file, which can greatly skew the results.
>
> This patch uses lseek to periodically restart at the beginning of the file,
> rather than writing past the end of it.

Oh, I never noticed that the later tests kept appending to the file
rather then overwriting it. I have applied the attached fix for CVS
HEAD that just uses lseek() before each write group, as you suggested.
I have backpatched it to 8.4.X because the original code created 16GB
files in tests (yikes).

--
Bruce Momjian <bruce(at)momjian(dot)us> http://momjian.us
EnterpriseDB http://enterprisedb.com

+ If your life is a hard drive, Christ can be your backup. +

Attachment Content-Type Size
/rtmp/diff text/x-diff 3.4 KB

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Tom Lane 2009-09-21 20:33:32 Re: Adding \ev view editor?
Previous Message Tom Lane 2009-09-21 20:11:59 Re: generic copy options