On Sun, May 05, 2019 at 08:37:02PM -0700, Daniel Farina wrote:
> Regrettably, it's too late, but I'll have my notes for next time.
Here is a guess: the base backup which has been used for this
instance's recovery has been taken from a standby, where the
minimum recovery LSN registered in its control file did not actually
map to the last physical LSN where a page has been flushed by either
the checkpointer, the startup process or a read-only backend.
This reminds me of the problems between 3c64dcb and c186ba1 I was at
the origin of, but this refers to 11.1, which was released after the
fact. Or perhaps a beta version of 11 has been used?
--
Michael