Re: [Patch] New psql prompt substitution %r (m = master, r = replica)

On Fri, Aug 12, 2016 at 12:51 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Peter Eisentraut <peter.eisentraut@2ndquadrant.com> writes:
>> On 8/11/16 6:04 AM, Aleksander Alekseev wrote:
>>> Suggested patch introduces an %r substitution in psql's prompt. This
>>> substitution allows to display whether user is connected to master or
>>> replica right in a prompt.
>
>> In the near future, there will (probably) be a lot more variants about
>> what it means to be a master or a replica.  There will be logical
>> replication, where you could be a publisher of something and a consumer
>> of something else.  You could even be a logical consumer but a physical
>> master.  So a global binary facility is probably not very forward
>> looking and will lead to confusion.
>
> Also, the patch as given is broken since it fails to account for the
> server being promoted while a psql session is open.
    {
+        {"is_master", PGC_INTERNAL, UNGROUPED,
+            gettext_noop("Shows whether the current instance is
master or replica."),
+            NULL,
+            GUC_REPORT | GUC_NO_SHOW_ALL | GUC_NO_RESET_ALL |
GUC_NOT_IN_SAMPLE | GUC_DISALLOW_IN_FILE
+        },
+        &session_is_master,
+        false,
+        NULL, NULL, NULL
+    },
Having a GUC for that purpose is not a fruitful approach. And it seems
to me that this patch is dead-in-the water because this makes prompt
parsing dependent on a status only known by the server, which would
require in the worst case to issue an SQL based on for example
pg_is_in_recovery() in get_prompt(): all the other fields using libpq
routines fetch values defined when the connection is established, like
the session PID or the database.
-- 
Michael