Обсуждение: duplicate key violate error

Hi,

I got duplicate key violate error in the db log for the following query:
INSERT INTO tab1 ( SELECT '1611576', '1187865' WHERE NOT EXISTS (        SELECT 1 FROM tab1         WHERE id='1611576' AND id2='1187865'    )    )

The error occured during production time.
But when I manually executed the query, it inserted one row with success and next time it inserted 0 rows.

\d tab1
id int
id2 int
primary key (id,id2)

So, any idea why the error occurred at production time.

AI Rumman wrote:
> I got duplicate key violate error in the db log for the following
query:
> INSERT INTO tab1 ( SELECT '1611576', '1187865' WHERE NOT EXISTS (
SELECT 1 FROM tab1
> WHERE id='1611576' AND id2='1187865'    )    )
>
> The error occured during production time.
> But when I manually executed the query, it inserted one row with
success and next time it inserted 0
> rows.
>
> \d tab1
> id int
> id2 int
> primary key (id,id2)
>
> So, any idea why the error occurred at production time.

Concurrency?

Session 1:

CREATE TABLE tab1 (
  id integer NOT NULL,
  id2 integer NOT NULL,
  PRIMARY KEY (id, id2)
);

START TRANSACTION;

INSERT INTO tab1
  (SELECT '1611576', '1187865' WHERE NOT EXISTS
    (SELECT 1 FROM tab1 WHERE id='1611576' AND id2='1187865')
  );

Session 2:

START TRANSACTION;

INSERT INTO tab1
  (SELECT '1611576', '1187865' WHERE NOT EXISTS
    (SELECT 1 FROM tab1 WHERE id='1611576' AND id2='1187865')
  );

Session 1:

COMMIT;

Session 2:

ERROR:  duplicate key value violates unique constraint "tab1_pkey"


Yours,
Laurenz Albe

On Tue, Jun 14, 2011 at 3:30 AM, AI Rumman <rummandba@gmail.com> wrote:
> Hi,
>
> I got duplicate key violate error in the db log for the following query:
> INSERT INTO tab1 ( SELECT '1611576', '1187865' WHERE NOT EXISTS (
> SELECT 1 FROM tab1         WHERE id='1611576' AND id2='1187865'    )    )
>
> The error occured during production time.
> But when I manually executed the query, it inserted one row with success and
> next time it inserted 0 rows.

Unfortunately the operation above is not atomic.  This is a classic
concurrency problem that everyone has to deal with -- there is no way
at present to rely on a simple row level lock to prevent concurrent
inserts to the same key.  You have a few of ways to deal with this:

*) retry the statement (personally not a big fan of this method)
*) lock the table (lousy concurrency)
*) advisory lock might work, if you must have concurrency and your key
is an integer. be careful, and do not overuse the technique.

merlin