On 05/14/2018 02:22 PM, Tom Lane wrote:
Adrian Klaver <adrian.klaver@xxxxxxxxxxx> writes:
On 05/14/2018 02:02 PM, Tom Lane wrote:
I didn't bother with spelling it all out in full detail this time,
which maybe was a mistake, but I felt that probably most users
wouldn't need to bother with these changes at all (unlike the case
where a catalog correction is security-related).
Well what is nice about the news release is you can cut and past the
entire list of commands and do the updates en masse.
It'd be nice to have some more-automated way of doing this type of
correction. Ordinary scripting doesn't look very promising, because
I don't see an easy way to deal with the need to connect to every
database in the cluster; that seems to depend on a lot of local
characteristics about usernames and authentication >
Maybe it'd be worth building some sort of infrastructure that would
allow this to be done at a lower level. It's not hard to imagine
an autovacuum-like or bgworker-based thingy that could run around
and apply a given SQL script in every database, bypassing the usual
worries about authentication and connections-disabled databases.
That seems like a lot of work for a need that only comes up once in
awhile, but perhaps it'd have more applications than just catalog
corrections.
That would be helpful given that a major version has a 5 year supported
life span. I can see folks not deciding to do the manual work at the
minor release because at that time it does not apply and the work does
not seem worth it. Then at some point in the future conditions change
and they wonder why things are not working the way they should. I know I
would be grateful.
regards, tom lane
--
Adrian Klaver
adrian.klaver@xxxxxxxxxxx
