I am using FreeBSD-4.4-STABLE, cvsup'd last weekend, and the version
is 7.1.3 as reflected in the FreeBSD port system.
I can try with 7.2beta, but it might be awkward as I only have a
dial-up line... if somebody with 7.2 already installed could try this,
you will almost certainly beat me to results.
I am interested if this occurs on Linux or other OSes so I can get
feedback to the FreeBSD port maintainer or OS if it is OS specific.
It looks like there is some semaphore contention, and the backends
call 'select' with a timeout AFTER the semaphore is let go. Runs like
10x slower than 1 of these tests when I run 2.
- Mike H.
Cc: pgsql-general@postgresql.org
Comments: In-reply-to Mike Harding <mvh@ix.netcom.com>
message dated "Sun, 25 Nov 2001 20:54:28 -0800"
Date: Mon, 26 Nov 2001 10:18:58 -0500
From: Tom Lane <tgl@sss.pgh.pa.us>
X-SpamBouncer: 1.4 (8/24/01)
X-SBClass: OK
Mike Harding <mvh@ix.netcom.com> writes:
> I am running a capacity test just for fun, and notice that if I run
> one update the system uses 100% of the CPU, but if I use 2 updates at
> the same time the system is only about 3% loaded and the postgres
> backends rotate between
Would you try your test case with 7.2beta?
BTW, you really should mention which version you're using, and be
more specific about the platform than "Freebsd".
regards, tom lane
