On Tue, Aug 11, 2015 at 1:12 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Maciek Sakrejda <maciek@heroku.com> writes:
>> On Mon, Aug 10, 2015 at 8:18 PM, Michael Paquier <michael.paquier@gmail.com>
>> wrote:
>>> But this fixed a problem on master where queries got stuck because of
>>> the WAL insert lock patch that has been introduced in 9.4 because of a
>>> race condition between two backends.
>
>> Interesting, thanks for pointing this out.
>
> FWIW, I doubt that that particular fix is related, because in a replica
> server there are no processes performing WAL inserts.
Yes, I was more thinking about a problem where XLogFlush interacts
badly with heap_delete. But that's hard to guess anything without
looking at the backtrace and a reproducible test case...
> A stack trace of the startup process would definitely be a good thing to
> grab if/when you see this again.
Yep.
--
Michael