Database system identifier via SELECT

From: Craig Ringer <craig(at)2ndquadrant(dot)com>
To: PostgreSQL Hackers <pgsql-hackers(at)postgresql(dot)org>
Cc: Peter Eisentraut <peter_e(at)gmx(dot)net>, chris(at)replicon(dot)com
Subject: Database system identifier via SELECT
Date: 2013-02-27 23:57:34
Message-ID: 512E9D6E.3000207@2ndquadrant.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Regarding these prior threads:

http://www.postgresql.org/message-id/CAC2SuRJLMtKyvCs6DbxdyUrHy+WKe-94cBuGnwCXAg914cwG7g@mail.gmail.com
http://www.postgresql.org/message-id/200707101624.47885.peter_e@gmx.net

It's somewhat less than pretty, but you *can* simply:

select substring(pg_read_binary_file('global/pg_control'),1,8);

then, knowing whether the system is little- or big-endian, decode the
resulting hex representation of a uint64 however you find to be
convenient, like say:

#!/usr/bin/python
import struct
import sys
print struct.unpack("Q", sys.argv[1].decode("hex"))[0]

... or your preferred spelling in C/Perl/LOLCode/whatever.

For many purposes the raw hex representation will be sufficient in any case.

The main downside here is the requirement of a superuser connection.

It'd be fairly trivial (and a lot saner) to write an extension to expose
GetSystemIdentifier() via SQL. Unfortunately that won't work for my
purposes because I need something that'll work on potentially quite old
servers and with minimal intrusiveness, but it'd be a reasonable answer
for people who want it for cluster/replication management.

--
Craig Ringer http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services

Browse pgsql-hackers by date

  From Date Subject
Next Message Peter Eisentraut 2013-02-28 03:16:00 issue with cpluspluscheck and systemtap
Previous Message Boszormenyi Zoltan 2013-02-27 19:38:29 Re: Strange Windows problem, lock_timeout test request