On Thu, 1 Sep 2011, Norman Yamada wrote:

> As was discussed in a previous thread here, pg_upgrade complains that the following slony tables/columns can't be migrated from 8.3.x to 9.0.x:

Also read over http://bugs.slony.info/bugzilla/show_bug.cgi?id=234
In addition to the issues you have discussed you will also have to make 
sure that the 64 bit version of the xid's post upgrade are newer than the 64 
bit versions after.  You might not see this issue on a new test cluster, 
you would only see it on a cluster that has had the 32 bit xid wrap around 
at least once.

An upgrade from 8.3 to 9.0 also will require you to make sure that the 8.4 
version of the slony stored functions are installed not the 8.3 one.

I have not tried upgrading a slony cluster with pg_upgrade, so there might 
be additional issues still.


>
>> sl_table.tab_relname
>> sl_table.tab_nspname
>> sl_table.tab_idxname
>> sl_sequence.seq_relname
>> sl_sequence.seq_nspname
>> vactables.relname
>
> I'd still like to use pg_upgrade on a large 8.3.x database that uses slony 2.0.7. if I do the following actions:
>
> 1) stop slony;
> 2) alter table sl_table alter column tab_relname type text, alter column tab_nspname type text, alter column tab_idxname type text;
>    alter table sl_sequence alter column seq_relname type text, alter column seq_nspname type text;
>    drop type vactables cascade;
>    CREATE TYPE vactables AS (
>        nspname text,
>        relname text
>   );
>   create function tablestovacuum () … leaving out function body…
>
> 3) pg_upgrade
>
> 4) start 9.0, do vacuuming analyze, etc…
>
> 5) fix slony back:
>
>  alter table sl_table alter column tab_relname type name, alter column tab_nspname type name, alter column tab_idxname type name;
>    alter table sl_sequence alter column seq_relname type name, alter column seq_nspname type name;
>    drop type vactables cascade;
>    CREATE TYPE vactables AS (
>        nspname name,
>        relname name
>   );
>   create function tablestovacuum () … leaving out function body…
>
> 6) then start slony
>
> Does anyone know if this works? I'm going to try this on a test cluster tomorrow, but wondering if anyone has done this already.
>
> Thanks,
>
> Norman Yamada
>
>
> _______________________________________________
> Slony1-general mailing list
> Slony1-general at lists.slony.info
> http://lists.slony.info/mailman/listinfo/slony1-general
>