Hi,
I have problem with deadlocks and don't know why it happens. Below is
the log (postgres 9.0, debian):

[11882]DETAIL: Process 11882 waits for ShareLock on transaction
44324308; blocked by process 11884.
Process 11884 waits for ShareLock on transaction 44324307;
blocked by process 11882.

Process 11882: SELECT * FROM teddy WHERE id IN
(112747007,112747008,112747011,112747013,112747015 ,112747016,112747020,112747021,112747022,112747024 ,112747025,112747028,112747030,112747032,112747034 ,112747035,112747038,112747043,112747044,112747045 ,112747050,112747052,112747053)
FOR UPDATE

Process 11884: SELECT * FROM teddy WHERE id IN
(112747007,112747008,112747011,112747013,112747015 ,112747016,112747020,112747021,112747022,112747024 ,112747025,112747028,112747030,112747032,112747034 ,112747035,112747038,112747043,112747044,112747045 ,112747050,112747052,112747053)
FOR UPDATE


Regards,
Roman

Roman <roman.janowczyk (AT) gmail (DOT) com> wrote:
The deadlock can happen because the two processes can lock the rows in
a different order.

In practice I believe if you add an ORDER BY to the SELECT FOR UPDATE
queries the locks will be taken in a consistent order (though this
doesn't appear to be a documented guarantee).

-M-

On Mon, 21 Feb 2011 19:02:18 +0000, Matthew Woodcraft wrote:

Deadlocks are also known as circular locks. Locking records in the same
order is a guarantee against deadlocks. That's not a software guarantee,
that's mathematics.



--
http://mgogala.byethost5.com