Am Dienstag, 19. August 2003 11:52 schrieb Philipp Reisner:
> Am Montag, 18. August 2003 15:38 schrieb Tom Lane:
> > Philipp Reisner <philipp.reisner@linbit.com> writes:
> > > Now if the applications issues one delete statement concurrently on
> > > two connections both block forever.
> > >
> > > Please correct me if I am wrong, but should not one of the statements
> > > succeed and the other simply fail ?
> > >
> > > Aug 18 10:34:25 nut1 postgres[4934]: [44389] LOG:
> > > query: delete from Calls where objID = 2268645
> > >
> > > Aug 18 10:34:25 nut1 postgres[4933]: [44071] LOG:
> > > query: delete from Calls where objID = 2268645
> > >
> > > 4933 ? S 5:20 postgres: sd sd 10.2.1.5 idle in transaction
> > > 4934 ? S 5:08 postgres: sd sd 10.2.1.5 DELETE waiting
> >
> > I see no deadlock. 4933 is waiting for its client to issue another
> > command. 4934 is waiting to see if it can delete the row.
> >
> > regards, tom lane
>
> Hi Tom,
>
> Right, thanks for pointing it out for me. It seems to be somwhere in
> the application code or PostgreSQL's jdbc driver.
>
> Finally we are able to reproduce the "Lockup" and we will continue
> to work out if it is in the App or in the jdbc driver.
>
> -Philipp
Just in order to end this thread. It turned out that the lockup was
triggered by an old version of postgres' jdbc driver. (It was from
the 7.2.1 days of postgresql)
Updating to a recent version finaly solved this issue.
-Philipp
