Re: [HACKERS] Possibly too stringent Assert() in b-tree code

Kyotaro HORIGUCHI <horiguchi(dot)kyotaro(at)lab(dot)ntt(dot)co(dot)jp> writes:
> At Mon, 26 Sep 2016 09:12:04 +0530, Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote in <CAA4eK1K5YyDmndko0zzW6WNCN_DGFVHa6DCYcyuvkBWTH5+nUQ(at)mail(dot)gmail(dot)com>
>>> It seems to me that we do take actions for conflict resolution during
>>> the page deletion (that looks to be covered by XLOG_HEAP2_CLEANUP_INFO
>>> which we emit in vacuum), but not sure if that is sufficient.
>>> Consider a case where the new transaction is started on standby after
>>
>> Here by new transaction, I intend to say some newer snapshot with
>> valid MyPgXact->xmin.

> I agree to the diagnosis. So the WAL record is not necessary if
> it is a new page since no one cannot be grabbing it.

Thanks for reviving this thread and reviewing the problem.
I pushed the patch now with some more work on the comments.

regards, tom lane