Re: pgsql: Reorder EPQ work, to fix rowmark related bugs and improve effici

Hi,

On September 9, 2019 11:01:28 PM GMT+01:00, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>Andres Freund <andres(at)anarazel(dot)de> writes:
>> On September 9, 2019 10:51:26 PM GMT+01:00, Thomas Munro
><thomas(dot)munro(at)gmail(dot)com> wrote:
>>> On Tue, Sep 10, 2019 at 12:32 AM Andres Freund <andres(at)anarazel(dot)de>
>>> Did this cause the following failure in eval-plan-qual on
>>> REL_12_STABLE?
>>>
>https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=prairiedog&dt=2019-09-09%2019%3A08%3A41
>
>Just saw that :-(
>
>> Argh, yes, probably. But it's just isolationtester I think, not a
>backed bug. Need to backport the recent fix in master to at least 12.
>> Tom, do you think we should backpatch both the order fix and the
>notice improvement, or just the former? And to which version?
>
>I think the first question is why is only prairiedog showing this.

We only saw the similar failure on master on a very small number of machines as well. Given The somewhat unusual way packets have to go out to be likely to be problematic, that's not too surprising.

>Let's not panic until we understand that.

If it's the packet ordering thing in isolationtester, then I think there's no need to panic, personally. And the output does look a lot like that.

>prairiedog looks to be about 20 minutes away from finishing the
>isolation tests for master, so that will be a useful data point
>as to whether there's any 12/HEAD difference. I'll dig deeper
>as soon as that run finishes.

Cool.

Andres
--
Sent from my Android device with K-9 Mail. Please excuse my brevity.