On Thu, Apr 7, 2022 at 4:03 PM kuroda.hayato@fujitsu.com
<kuroda.hayato@fujitsu.com> wrote:
>
> Dear Peter,
>
> > FYI, here is a test script that is using the current patch (v6) to
> > demonstrate a way to share table data between different numbers of
> > nodes (up to 5 of them here).
>
> Thanks for sharing your script! It's very helpful for us.
>
> While reading your script, however, I had a question about it.
> Line 121-122, you defined subscriptions for 2-nodes cluster:
>
> psql -p $port_N1 -c "create subscription sub12 connection 'port=$port_N2' publication pub2 with ($copy_force);"
> psql -p $port_N2 -c "create subscription sub21 connection 'port=$port_N1' publication pub1 with ($copy_force);"
>
> But I was not sure it works well.
> N2 already have shared data from N1 when subscription sub21 is created.
> Did you assume that the initial copying is not so quick and
> data synchronization will be not done when creating sub21?
Oops. Good catch.
Although the 2-way test was working OK for me, I think that it worked
only because of lucky timing. e.g. When I put a delay between those 2
subscriptions then the 2nd one would cause the PK violation that
probably you were anticipating would happen.
I have modified the 2-way example to use the same truncate pattern as others.
PSA the fixed test.sh script and accompanying files.
------
Kind Regards,
Peter Smith.
Fujitsu Australia
