My mistake. The index on pg_shadow breaks nearly everything. I can
have triggers maintain this effectively enough for my needs however.
However, whether or not the number of users I want to add is going to
be too much is still a question.
--
Rod Taylor
There are always four sides to every story: your side, their side, the
truth, and what really happened.
----- Original Message -----
From: "Rod Taylor" <rod.taylor@inquent.com>
To: "Hackers List" <pgsql-hackers@postgresql.org>
Sent: Tuesday, January 30, 2001 2:24 PM
Subject: [HACKERS] pg_shadow overusage
> I'd like to keep a complete history of everything that users of the
> system does from the database level (as there are a number of
> applications, and backend workers). I've done this through various
> rules, triggers.
>
> The next step is to try to tie a name to it. getpgusername()
supplies
> that perfectly. The real question is, how many users can be in
> pg_shadow without the system thrashing.
>
> We're looking to start with approx. 20 000, and will likely grow
that
> number to 100k users in a years time (I understand a goal for 7.2 is
> to get rid of the OID limit :).
>
> Will pg_shadow and the database in general barf with that number of
> users in the database? A limit of 40 simultaneous connections
should
> suffice for our needs as user sessions will be quite short.
>
> I also want to FOREIGN KEY the usename column in pg_shadow, I'll
> assume that adding a unique index won't hurt anything (as I've not
> noticed any problems on the test system thus far).
>
> Thanks for any input.
>
> --
> Rod Taylor
>
> There are always four sides to every story: your side, their side,
the
> truth, and what really happened.
>
