Andrew Dunstan <andrew@dunslane.net> writes:
> Tom Lane wrote:
>> ... and do what?
> In general, I think I'd probably prefer normal database startup to fail
> if a tablespace is missing. That way I will know about it right then and
> can remedy it. This is something that is much more likely to happen than
> an arbitrary missing file, ISTM, and at the same time the check would be
> much easier than looking for an arbitrary file. At the very least we
> could issue a warning.
So what you're imagining is
* iterate through each symlink in $PGDATA/pg_tblspc
* check that PG_VERSION exists (and has the right contents??) in each pointed-to directory
* fail if not
I guess this is reasonable, since we make a similar check for the core
data directory itself. Watch out for the initdb sequence though.
regards, tom lane