Vladimir Borodin wrote:
> I’m not sure that these 104 bytes will always be 104 bytes to have a
> strict equality while checking. Could it change in the future?
There is no promise that WAL record format stays unchanged. Sometimes
we change a WAL record in a minor release.
> Or is there a better way to understand that streaming replica received
> all data after master shutdown? The check that pg_xlog_location_diff
> returns 104 bytes seems a bit strange.
I guess you could pg_xlogdump the difference and verify that it is a
shutdown checkpoint record. As far as I remember there should always be
one at the end of recovery.
--
Álvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services