Re: More info

From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: Tim Perdue <tperdue(at)valinux(dot)com>
Cc: pgsql-hackers(at)hub(dot)org
Subject: Re: More info
Date: 2000-07-10 04:39:09
Message-ID: 24935.963203949@sss.pgh.pa.us
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Tim Perdue <tperdue(at)valinux(dot)com> writes:
> Any idea why 6.5.3 would have created tens of thousands of files like
> these in the /data/base/db_geocrawler/ directory?

Funny you should mention that, because I was just in process of testing
a fix when your mail came in. The low-level routine that accesses a
particular segment of a multi-segment relation develops a serious case
of Sorcerer's Apprentice syndrome if higher levels hand it a silly block
number. If you tell it to access, say, block# 2 billion, it will
merrily start creating empty segment files till it gets to the segment
number that corresponds to that block number.

The routine does need to be able to create *one* new segment, in case it
is asked to access the block just past the current EOF (when EOF is at a
segment boundary) ... but not more than one. As of current sources, it
knows not to do more.

This bug has been known for a while. It doesn't directly answer your
problem though, since the real issue is "what generated the silly block
number, and why"?

I can't quite resist the temptation to suggest that you should be
running 7.0.2 ...

regards, tom lane

In response to

  • More info at 2000-07-10 03:52:47 from Tim Perdue

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Tom Lane 2000-07-10 05:02:19 Re: Re: [GENERAL] PostgreSQL vs. MySQL
Previous Message Graeme Merrall 2000-07-10 04:31:52 Re: Slashdot discussion