On Tue, May 07, 2019 at 10:31:59AM +0900, Kyotaro HORIGUCHI wrote:
> The patch seems to be using the tablespace directory in backups
> directly from standbys. In other words, multiple standbys created
> from A backup shares the tablespace directory in the backup.
Yes, I noticed that, and I am not happy about that either. I'd like
to think that what we are looking for is an equivalent of the
tablespace mapping of pg_basebackup, but for init_from_backup(). At
least that sounds like a plausible solution.
--
Michael