Re: [PERFORM] Strange performance degradation - Mailing list pgsql-general

From Matthew Wakeling
Subject Re: [PERFORM] Strange performance degradation
Date
Msg-id alpine.DEB.2.00.0911241527020.684@aragorn.flymine.org
Whole thread Raw
In response to Strange performance degradation  (Lorenzo Allegrucci <lorenzo.allegrucci@forinicom.it>)
Responses Re: [PERFORM] Strange performance degradation  (Denis Lussier <denis.lussier@enterprisedb.com>)
List pgsql-general
On Tue, 24 Nov 2009, Denis Lussier wrote:
> IMHO the client application is already confused and it's in Prod.
> Shouldn't he perhaps terminate/abort the IDLE connections in Prod and
> work on correcting the problem so it doesn't occur in Dev/Test??

The problem is, the connection isn't just IDLE - it is idle IN
TRANSACTION. This means that there is quite possibly some data that has
been modified in that transaction. If you kill the backend, then that will
automatically roll back the transaction, and all of those changes would be
lost.

I agree that correcting the problem in dev/test is the priority, but I
would be very cautious about killing transactions in production. You don't
know what data is uncommitted. The safest thing to do may be to bounce the
application, rather than Postgres.

Matthew

--
 All of this sounds mildly turgid and messy and confusing... but what the
 heck. That's what programming's all about, really
                                        -- Computer Science Lecturer

pgsql-general by date:

Previous
From: Lorenzo Allegrucci
Date:
Subject: Re: [PERFORM] Strange performance degradation
Next
From: Alexandra Roy
Date:
Subject: Re: DBD::Pg 2.15.1 compilation failed