On 2016-03-15 14:21:34 -0400, Robert Haas wrote:
> On Tue, Mar 8, 2016 at 6:08 PM, Andres Freund <andres@anarazel.de> wrote:
> > On 2016-03-08 16:42:37 -0500, Robert Haas wrote:
> >> - I really wonder if the decision to ignore sessions that are idle in
> >> transaction (aborted) should revisited. Consider this:
> >>
> >> rhaas=# begin;
> >> BEGIN
> >> rhaas=# lock table pg_class;
> >> LOCK TABLE
> >> rhaas=# savepoint a;
> >> SAVEPOINT
> >> rhaas=# select 1/0;
> >> ERROR: division by zero
> >
> > Probably only if the toplevel transaction is also aborted. Might not be
> > entirely trivial to determine.
>
> Yes, that would be one way to do it - or just ignore whether it's
> aborted or not and make the timeout always apply. That seems pretty
> reasonable, too, because a transaction that's idle in transaction and
> aborted could easily be one that the client has forgotten about, even
> if it's not hanging onto any resources other than a connection slot.
> And, if it turns out that the client didn't forget about it, well,
> they don't lose anything by retrying the transaction on a new
> connection anyway.
I'm fine with both.
Andres