Re: "ERROR: deadlock detected" when replicating TRUNCATE

On Mon, May 17, 2021 at 8:13 PM Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
>
> On Mon, May 17, 2021 at 3:05 PM Dilip Kumar <dilipbalaut(at)gmail(dot)com> wrote:
> >
> > On Mon, May 17, 2021 at 12:30 PM Peter Smith <smithpb2250(at)gmail(dot)com> wrote:
> >
> > > PSA a patch adding a test for this scenario.
> >
> > I am not sure this test case is exactly targeting the problematic
> > behavior because that will depends upon the order of execution of the
> > apply workers right?
> >
>
> Yeah, so we can't guarantee that this test will always reproduce the
> problem but OTOH, I have tried two times and it reproduced both times.
> I guess we don't have a similar test where Truncate will replicate to
> two subscriptions, otherwise, we would have caught such a problem.
> Having said that, I am fine with leaving this test if others feel so
> on the grounds that it won't always lead to the problem reported.
>

If there is any concern that the problem won't always happen then I
think we should just increase the numbers of subscriptions.

Having more simultaneous subscriptions (e.g. I have tried 4). will
make it much more likely for the test to encounter the deadlock, and
it probably would also be quite a useful worker stress test in it's
own right.

~~

Also, should this test be in the 010_truncate,pl, or does it belong in
the 100_bugs.pl? (I don't know what are the rules for when a test
gets put into 100_bugs.pl)

----------
Kind Regards,
Peter Smith.
Fujitsu Australia