On 9/2/19 4:57 PM, James Sewell wrote:
>
>
>
> I am going to assume then that it has to do with this:
> "LINE 39: "location_pt" "public"."geography"(Point,4283), "
>
> What is the above pointing to?
>
>
> This needs the PostGIS types and tables to work - they don't exist as
> they were not created with CREATE EXTENSION, but rather are to be
> created later in the restore.
>
>
>
> >
> > 2) "Both the versions have PostGIS 2.5.1 installed and working"
> >
> > How do you know it is working on the 11 version?
> >
> >
> > By version I mean 11 - I can init a new cluster and it’s fine
> >
> > The more I look at this it seems like:
> >
> > A) pg_upgrade somehow creates postgis without supporting tables
>
> So you see a CREATE EXTENSION postgis?
>
> Is there anything in the logs that pertains to the above?
>
>
> Yes and no. Just the create extension.
>
>
>
> > B) while the tables would be created from the schema dump files,
> they
> > happen too late
>
> What happens if you use pg_dump from version 11 against the 9.6 cluster
> with the -s and -p?
>
> Does it shows the schema creation in the order you need?
>
>
> Yes. It's only in the binary_upgrade mode I see issues (because the
> extension stuff isn't created with CREATE EXTENSION)
>
What is the pg_upgrade command you are using?
--
Adrian Klaver
adrian.klaver@aklaver.com