Thread: Increased BLKSZ, but now pgsql seg-faults?

Hello,

I've increased the BLKSZ to 32K to support large tuples and then I did a full
re-compile (gmake clean, configure, gmake all, gmake install) and the database
accepts my large tuples. But now when I come in through psql, I get a
segmentation fault inside of psql when I try to display the table with the large
tuples (i.e. \d largetable).

I know psql is for a user-interface to the database, and my application won't be
needing this in the end, but it would be nice to use psql for debugging and
development. Is there a way to increase the buffer sizes (I'm assuming that is
what is causing the problem). Here is the output when it seq-faults:

Segmentation Fault (core dumped)
pq_recvbuf: unexpected EOF on client connection

Wade Oberpriller

> I've increased the BLKSZ to 32K to support large tuples and then I did a
full
> re-compile (gmake clean, configure, gmake all, gmake install) and the
database
> accepts my large tuples. But now when I come in through psql, I get a
> segmentation fault inside of psql when I try to display the table with the
large
> tuples (i.e. \d largetable).

  The developpers have said several times here that increaseing BLKSZ has
bad side-effects. : )

steve

I set my BLCKSZ to 30720 (30k -- so as to not be absolute bleeding edge) and
everything has been running perfect. It's on my development machine but I
hit it hard for a straight week just to make sure it was going to be stable.
I just used some scripts of my own making to do enormous amounts of INSERT
UPDATE DELETE queries. I think over the week I did several hundred thousand
(it's been a few months back).. Still, it worked fine for me..

I see the following in config.h, have you read this?

/*
 * RELSEG_SIZE is the maximum number of blocks allowed in one disk file.
 * Thus, the maximum size of a single file is RELSEG_SIZE * BLCKSZ;
 * relations bigger than that are divided into multiple files.
 *
 * CAUTION: RELSEG_SIZE * BLCKSZ must be less than your OS' limit on file
 * size.  This is typically 2Gb or 4Gb in a 32-bit operating system.  By
 * default, we make the limit 1Gb to avoid any possible integer-overflow
 * problems within the OS.  A limit smaller than necessary only means we
 * divide a large relation into more chunks than necessary, so it seems
 * best to err in the direction of a small limit.  (Besides, a power-of-2
 * value saves a few cycles in md.c.)
 *
 * CAUTION: you had best do an initdb if you change either BLCKSZ or
 * RELSEG_SIZE.
 */


-Mitch


----- Original Message -----
From: "Steve Wolfe" <steve@iboats.com>
To: "general-help postgresql" <pgsql-general@postgresql.org>
Sent: Wednesday, September 27, 2000 11:12 AM
Subject: Re: [GENERAL] Increased BLKSZ, but now pgsql seg-faults?


> > I've increased the BLKSZ to 32K to support large tuples and then I did a
> full
> > re-compile (gmake clean, configure, gmake all, gmake install) and the
> database
> > accepts my large tuples. But now when I come in through psql, I get a
> > segmentation fault inside of psql when I try to display the table with
the
> large
> > tuples (i.e. \d largetable).
>
>   The developpers have said several times here that increaseing BLKSZ has
> bad side-effects. : )
>
> steve
>
>
>

oberpwd@anubis.network.com (Wade D. Oberpriller) writes:
> I've increased the BLKSZ to 32K to support large tuples and then I did
> a full re-compile (gmake clean, configure, gmake all, gmake install)
> and the database accepts my large tuples. But now when I come in
> through psql, I get a segmentation fault inside of psql when I try to
> display the table with the large tuples (i.e. \d largetable).

In psql?  That's pretty odd -- psql shouldn't even be able to determine
what blocksize the server is using, much less be affected by it.  I
suspect you may have unearthed an unrelated bug.

Would you (re)compile psql with -g and send a stack trace of the fault?

            regards, tom lane