Dying PostgreSQL backend

From: "otisg" <otisg(at)iVillage(dot)com>
To: <pgsql-admin(at)postgresql(dot)org>
Subject: Dying PostgreSQL backend
Date: 2002-05-06 05:01:08
Message-ID: 002501c1f4bb$087b8920$6ac8010a@mail2world.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-admin

Hello,

I am having problems keeping PostgreSQL alive in certain situations and
I wonder if I can tune some parameters to allow it to handle higher
load.

What gets logged in syslog log file are things like this:

May 5 21:52:34 localhost postgres[25012]: [64] NOTICE: TRANSACTION
ISOLATION LEVEL is READ COMMITTED
May 5 21:52:34 localhost postgres[25013]: [61-1] NOTICE: Message from
PostgreSQL backend:
May 5 21:52:34 localhost postgres[25013]: [61-2] ^IThe Postmaster has
informed me that some other backend
May 5 21:52:34 localhost postgres[25013]: [61-3] ^Idied abnormally and
possibly corrupted shared memory.
May 5 21:52:34 localhost postgres[25013]: [61-4] ^II have rolled back
the current transaction and am
May 5 21:52:34 localhost postgres[25013]: [61-5] ^Igoing to terminate
your database system connection and exit.
May 5 21:52:34 localhost postgres[25013]: [61-6] ^IPlease reconnect to
the database system and repeat your query.
May 5 21:52:34 localhost postgres[25015]: [61] FATAL 1: The database
system is starting up
...

The DB then restarts.

This happens when I hit the DB with a few simple UPDATE statements a
second.
The statements are really simple, changing 3 TIMESTAMP, 2 INT, and one
VARCHAR field using just the primary key to match the row to update.
I'm executing about 6-8 of those per second.
The DB I have has less than 10 tables, most of which are completely
empty, and just one, the one I'm updating has about 12,000 pretty narrow
(< 10 columns) rows.
I have run VACUUM ANALYZE on this DB, but once I start my application,
which seems to be able to do about 7000-8000 updates before postgreSQL
backend starts dying, I do not run VACUUM ANALYZE.

I have already changed a few postgresql.conf parameters:

shared_buffers = 64
sort_mem = 1024
fsync = false
deadlock_timeout = 5000

The UPDATE statements are the only ones running against this DB (except
for a simple, no join, straight forward SELECT that runs every 20
seconds).

What other parameters can I change?
Could the problem be that I'm doing a lot of UPDATEs without running
VACUUM?

Thanks,
Otis
_______________________________________________________________
Sign up for FREE iVillage newsletters <http://s.ivillage.com/rd/16705> .
>From health and pregnancy to shopping and relationships, iVillage
has the scoop on what matters most to you.

Responses

Browse pgsql-admin by date

  From Date Subject
Next Message Gaetano Mendola 2002-05-06 10:46:18 SERIAL Field
Previous Message Lamar Owen 2002-05-06 04:38:22 Re: Cannot (re)start PostgreSQL