Re: Mutable listen_addresses GUC

From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: Ivan Kovmir <ivan(dot)kovmir(at)cybertec(dot)at>
Cc: pgsql-hackers(at)postgresql(dot)org
Subject: Re: Mutable listen_addresses GUC
Date: 2025-09-29 14:43:35
Message-ID: 197013.1759157015@sss.pgh.pa.us
Views: Whole Thread | Raw Message | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Ivan Kovmir <ivan(dot)kovmir(at)cybertec(dot)at> writes:
> It is necessary to restart PostgreSQL to bind to a different network
> interface, thus breaking the active connections.

On most platforms, if you set listen_addresses to "*" then there's no
problem, the kernel will automatically cope with IP address changes.
Do we really need more than that?

The reason I'm pushing back is that this patch looks extremely
complicated and hard-to-test. It also violates one of the fundamental
precepts of the GUC subsystem, namely that assign hooks shalt not
fail. And there is a lot of stuff you've not covered, such as the
logic around whether to register for Bonjour. We could possibly
make this feature work, but the cost-benefit ratio looks quite poor.

regards, tom lane

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Rafia Sabih 2025-09-29 14:51:13 Re: Bypassing cursors in postgres_fdw to enable parallel plans
Previous Message Tom Lane 2025-09-29 14:27:57 Re: plan shape work