Обсуждение: PG crashes MacOS 10.3.4

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


I am having some backend stability problems with postgresql 7.4.3 on
MacOS 10.3.4 ($OSTYPE=darwin7.0). I compiled 7.4.3 from the source less
than two weeks ago.

At random intervals that seem to vary by database uptime and usage, a
COPY [table] FROM [file] statement involving maybe 20k - 80k rows
(larger amounts of data than I usually work with, but nothing the
server shouldn't be able to handle) will crash the machine. The screen
freezes, except for the mouse which freezes after about 30 seconds, and
the machine doesn't respond to keyboard or mouse input. Attempts to SSH
into the machine fail-- its not just the GUI locking up. The machine
must be cold-booted with the reset button. After the machine reboots,
the exact same COPY FROM statement can be executed without incident.
Nothing revealing is written to system.log, console.log, or the
postgresql logfile in debug mode.

I have confirmed it is the database server itself, and not the client
(which in this case happens to be DBD::Pg), which is crashing by
running the client on another machine.

I have not changed any of the defaults in postgresql.conf.

Any ideas?

- -Colin Stagner


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (Darwin)

iD8DBQFBArs/uq9VKK6yTRkRAo9RAJ0fIoZiPLiIDYO9H+j/IPBEJ2DF6wCdFvpo
DUypd+v5j5ggcW0SWfOL2fY=
=/Gm7
-----END PGP SIGNATURE-----

Colin Stagner <cbs228@earthlink.net> writes:
> At random intervals that seem to vary by database uptime and usage, a
> COPY [table] FROM [file] statement involving maybe 20k - 80k rows
> (larger amounts of data than I usually work with, but nothing the
> server shouldn't be able to handle) will crash the machine. The screen
> freezes, except for the mouse which freezes after about 30 seconds, and
> the machine doesn't respond to keyboard or mouse input. Attempts to SSH
> into the machine fail-- its not just the GUI locking up. The machine
> must be cold-booted with the reset button.

As an unprivileged process, Postgres *cannot* crash the machine --- at
least not on sanely-written operating systems, and OS X qualifies.  A
system-wide lockup is either a hardware fault or a kernel bug.

My money is on an intermittent hardware problem.

            regards, tom lane