On Mon, Sep 8, 2025 at 8:20 AM Chao Li <li.evan.c...@gmail.com> wrote:
>
> Hi Amit,
>
> Thanks for your info.
>
> On Sep 6, 2025, at 12:32, Amit Kapila <amit.kapil...@gmail.com> wrote:
>
> You can avoid this problem by creating a slot first on publisher with
> something like:
> postgres=# select pg_create_logical_replication_slot('s1', 'pgoutput',
> false, true);
> pg_create_logical_replication_slot
> ------------------------------------
> (s1,0/01BFF178)
> (1 row)
>
> Then while creating subscription you can use the above created slot as
> follows:
> db1=# create subscription sub1 connection 'dbname=postgres'
> publication pub1 WITH(create_slot=false, slot_name='s1');
> CREATE SUBSCRIPTION
>
> --
> With Regards,
> Amit Kapila.
>
>
> I am aware of this workaround solution. When I encountered the “create
> subscription” stuck problem, I did a Google search, then I found the old
> discussion thread that mentioned a workaround, so I did further search and
> eventually figured out the workaround.
>
> I think it’s worth resolving this issue, because it avoids confusion from
> people who want to try or test logical replication logically, and allows the
> same script to run in both an pure local environment and a real non-local
> environment. WDYT?
>
I don't see why someone can't use the workaround provided. The
solution proposed is more like adding hacks in the code making it a
maintenance burden.
--
With Regards,
Amit Kapila.
