From: | Stephan Szabo <sszabo(at)megazone23(dot)bigpanda(dot)com> |
---|---|
To: | Talja Ari <Ari(dot)Talja(at)comptel(dot)com> |
Cc: | "'pgsql-bugs(at)postgresql(dot)org'" <pgsql-bugs(at)postgresql(dot)org> |
Subject: | Re: now() gives the time of the last commit, not the time it |
Date: | 2002-10-23 13:16:08 |
Message-ID: | 20021023061500.B3876-100000@megazone23.bigpanda.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-bugs |
On Wed, 16 Oct 2002, Talja Ari wrote:
> Postgres version: 7.2.1
> Platform: Server on Linux, (java-)client on HP-UX.11 and SunOS 5.8
> JDBC drivers: For PostgreSQL 7.2, JDK 1.3
> JDK version: 1.3.1
> Autocommit off for each connection
> Special server flags: -i to allow the connections from different host
>
> The program is a java server, which updates a row in the database at regular
> intervals. The update time is taken from postgres function now() but the
> update time is the time when the connection was created/last transaction
> ended. In Oracle the time which is created by 'sysdate' is the time when the
> update statement is executed (or the time when the statement is actually
> committed. I'm not sure about that because the execution and commit are
> called practically at the same time). This causes a delay to the time and
> the size of the delay can be almost anything depending on the time how long
> the connection has been without any use. The connections are kept in a
> connection pool for later use so I cannot trust that the connection is
> created or the transaction has ended recently enough. Currently I'm calling
> commit before I execute the update statement.
This is the currently intended behavior of now(), if you want the moment
the call is made, you can use timeofday() I believe.
From | Date | Subject | |
---|---|---|---|
Next Message | Stephan Szabo | 2002-10-23 13:19:50 | Re: vacuum bug |
Previous Message | Robert Hentosh | 2002-10-22 22:38:12 | Re: Regression test failure. |