From: Andrew Gierth <andrew@tao11.riddles.org.uk>
To: "S. Bob" <sbob@quadratum-braccas.com>
Cc: pgsql-admin <pgsql-admin@postgresql.org>
Subject: Re: Unique cluster / instance identifier?
In-Reply-To: <92d53982-cf1d-ce3f-c541-60a252f85aef@quadratum-braccas.com> (S.
	Bob's message of "Mon, 18 Feb 2019 10:08:11 -0700")
Message-ID: <87r2c5htm3.fsf@news-spur.riddles.org.uk>
References: <92d53982-cf1d-ce3f-c541-60a252f85aef@quadratum-braccas.com>
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/25.1 (berkeley-unix)
Date: Mon, 18 Feb 2019 18:03:34 +0000
MIME-Version: 1.0
Content-Type: text/plain
Precedence: bulk

>>>>> "S" == S Bob <sbob@quadratum-braccas.com> writes:

 S> Hi all;

 S> Does PostgreSQL store a unique cluster or instance identifier that
 S> we can access to identify our instances?

Yes. But it's not easy to get at: it's the "Database system identifier"
shown in the output of pg_controldata. (pg12+ will have a function to
get it from SQL, but released versions don't.)

e.g.

Database system identifier:           6659016342798875639

The number is mostly generated from the time the instance was created,
with some bits modified. So for example:

$ perl -le 'print scalar gmtime +(shift >> 32)' 6659016342798875639
Sun Feb 17 17:04:21 2019

PG uses this internally to make sure that replication never tries to
apply WAL records generated on one system to a different system.
Physical replication secondary systems share the primary's system
identifier since they always originate as a copy (via base backup) of
the primary.

-- 
Andrew (irc:RhodiumToad)