On 2015/09/02 20:42, Amit Kapila wrote:
> On Wed, Sep 2, 2015 at 4:19 PM, Pavan Deolasee <pavan.deolasee@gmail.com
> <mailto:pavan.deolasee@gmail.com>> wrote:
> > On Wed, Sep 2, 2015 at 3:55 PM, Amit Langote
> <Langote_Amit_f8@lab.ntt.co.jp <mailto:Langote_Amit_f8@lab.ntt.co.jp>>
> wrote:
> >> On 2015-09-02 PM 06:41, Amit Langote wrote:
> >> Which, I now realize, is not the worry Amit Kapila's expresses.
> >> The deadlock was *indeed detected* in this case, with all the locks
> in the
> >> same PG instance. In a sharded environment with multiple PG instances,
> >> that becomes tricky. DLM (distributed lock manager/deadlock detector)
> >> seems indeed necessary as Amit K. suspects.
Ah, you are right.
> > Right. XC/XL did not address this issue and they rely on statement
> timeouts to break distributed deadlocks.
> I think that will be difficult for application to decide and then it
> needs to decide the same for all statements which is tricky because
> different statements could take different time. I think it is better to
> have solution for this problem and deadlock should be detected.
+1
Best regards,
Etsuro Fujita