Michael Paquier <mich...@paquier.xyz> writes:
> On Thu, Apr 18, 2019 at 05:57:39PM -0400, Tom Lane wrote:
>> Anyway, this is *not* new in v12.
> Indeed. It seems to me that v12 makes the problem easier to appear
> though, and I got to wonder if c6c9474 is helping in that as more
> cases are popping up since mid-March.
Yeah. Whether that's due to a server code change, or new or modified
test cases, is unknown. But looking at my summary of buildfarm runs
that failed like this, there's a really clear breakpoint at
gull | 2018-08-24 03:27:16 | recoveryCheck | pg_ctl: server
does not shut down
Since that, most of the failures with this message have been in the
recoveryCheck step. Before that, the failures were all over the
place, and now that I look closely a big fraction of them were
in bursts on particular animals, suggesting it was more about
some local problem on that animal than any real code issue.
So it might be worth groveling around in the commit logs from
last August...
regards, tom lane
