We have increased shared buffers to 2.5GB (linux kernels allows us reach this level) and lowered work_mem to 500MB.
Let's see tonight. :-)
-----Original Message-----
From: Tom Lane <tgl@xxxxxxxxxxxxx>
To: Iñigo Martinez Lasala <imartinez@xxxxxxxxxxxx>
Cc: pgsql-admin@xxxxxxxxxxxxxx
Subject: Re: Log full with gigabyes of CurTransactionContex
Date: Mon, 15 Jun 2009 10:02:26 -0400
=?ISO-8859-1?Q?I=F1igo?= Martinez Lasala <imartinez@xxxxxxxxxxxx> writes: > We have a problem with an insert query in one of our clients. This query > is launched in a night batch process. We have observed that if change > set is big (more than 50.000 updates) our database log starts growing > with thousands and thousands of lines until the server is out of space > and database freezes. You're running out of memory. > work_mem = 2000MB > maintenance_work_mem = 1024M These two settings are probably the cause. With shared_buffers at 2GB, you do not have anywhere near 1GB to play around with in a 32-bit environment. Try something like 200M and 500M. > Increasing temp buffers could help? I can hardly think of anything more counterproductive. You don't have enough address space. regards, tom lane
