Re: Database transaction related

could you please elaborate this concept of queue table?
~Jas

On 5/12/07, Martijn van Oosterhout <kleptog(at)svana(dot)org> wrote:
>
> On Fri, May 11, 2007 at 04:07:10PM -0400, Jasbinder Singh Bali wrote:
> > I have a transaction in postgres database as follows:
> >
> > 1 Insert in table xyz
> > 2 trigger fired on table xyz after insert
> > 3 Trigger calls a function func
> > 4 func is an implementation of a client socket in perl
> >
> > 1-4 happens in database environment only
>
> <snip>
>
> > Now my question is, what all happens in the unix tools server, is that
> > a part of the database transaction that started from step 1 above?
>
> No, how could the database know they are the same transaction? You
> would somehow have to get the unix server to send commands via your
> socket.
>
> What you're doing is usually the wrong approach. What people usually do
> is have the trigger insert a row into a queue table and have the unix
> tools server connect and do the work listed in the table, possibly
> marking the original row "done" somehow.
>
> > Why I'm concerned about all this is because off late, a database
> > insert failed in the unix tool server and the whole transaction start
> > from step 1 above was not rolled back. It was still successful till
> > step 4.
>
> To make that work the client would have to detect the unix tools server
> failed and abort the transaction locally also...
>
> Have a nice day,
> --
> Martijn van Oosterhout <kleptog(at)svana(dot)org> http://svana.org/kleptog/
> > From each according to his ability. To each according to his ability to
> litigate.
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.1 (GNU/Linux)
>
> iD8DBQFGRYxoIB7bNG8LQkwRAl5uAJ998zJyTN9S48bPYm0nM8qMW5D5DgCfQAXc
> tnDynaKd9KXyatpWGDkYDMw=
> =IXZO
> -----END PGP SIGNATURE-----
>
>