Re: pg_upgrade does not upgrade pg_stat_statements properly - Mailing list pgsql-hackers

From Bruce Momjian
Subject Re: pg_upgrade does not upgrade pg_stat_statements properly
Date
Msg-id 20210729161432.GN9600@momjian.us
Whole thread Raw
In response to Re: pg_upgrade does not upgrade pg_stat_statements properly  (Jan Wieck <jan@wi3ck.info>)
Responses Re: pg_upgrade does not upgrade pg_stat_statements properly
List pgsql-hackers
On Thu, Jul 29, 2021 at 11:46:12AM -0400, Jan Wieck wrote:
> On 7/29/21 11:10 AM, Bruce Momjian wrote:
> > On Thu, Jul 29, 2021 at 11:01:43AM -0400, Dave Cramer wrote:
> > > Much better, however I'm unclear on whether CREATE EXTENSION is actually
> > > executed on the upgraded server.
> > > 
> > > From what I could gather it is not, otherwise the new function definitions
> > > should have been in place. I think what happens is that the function
> > > definitions are copied as part of the
> > > schema and pg_extension is also copied.
> > 
> > Yes, the _effect_ of CREATE EXTENSION in the old cluster is copied to
> > the new cluster as object definitions.  CREATE EXTENSION runs the SQL
> > files associated with the extension.
> > 
> 
> This assumes that the scripts executed during CREATE EXTENSION have no
> conditional code in them that depends on the server version. Running the
> same SQL script on different server versions can have different effects.
> 
> I don't have a ready example of such an extension, but if we ever would
> convert the backend parts of Slony into an extension, it would be one.

The bottom line is that we have _no_ mechanism to handle this except
uninstalling the extension before the upgrade and re-installing it
afterward, which isn't clearly spelled out for each extension, as far as
I know, and would not work for extensions that create data types.

Yes, I do feel this is a big hold in our upgrade instructions.

-- 
  Bruce Momjian  <bruce@momjian.us>        https://momjian.us
  EDB                                      https://enterprisedb.com

  If only the physical world exists, free will is an illusion.




pgsql-hackers by date:

Previous
From: Bruce Momjian
Date:
Subject: Re: pg_upgrade does not upgrade pg_stat_statements properly
Next
From: Robert Haas
Date:
Subject: Re: [Patch] ALTER SYSTEM READ ONLY