Andres Freund wrote:
> On 2017-12-26 13:31:03 -0300, Alvaro Herrera wrote:
> > It's strange that this has gone undetected for so long. I wonder if
> > there's an interaction with logical decoding and its historical
> > snapshot stuff here.
>
> I can't see how - did you have a vague theory you could share?
Not really. I was just looking at the snapmgr.c code and thinking
whether there was a chance that it was the historic snapshot could be
involved in waits. But then, as presented the failure does not use
logical decoding, so even if there's a bug there (for which I have no
evidence) it would have to be a separate one.
More generally I was wondering if there was anyplace else that would
keep a catalog snapshot registered for a long time and then go to sleep.
Looked for it desultorily, came up with nothing.
Thanks,
--
Álvaro Herrera https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services