From: Tom Lane <tgl@sss.pgh.pa.us>
To: Murthy Nunna <mnunna@fnal.gov>
cc: "'pgsql-admin@postgresql.org'" <pgsql-admin@postgresql.org>
Subject: Re: Replication Question
In-reply-to: 
 <DM6PR09MB3386282980E4FDF4429A0A18B87F0@DM6PR09MB3386.namprd09.prod.outlook.com>
References: 
 <DM6PR09MB3386282980E4FDF4429A0A18B87F0@DM6PR09MB3386.namprd09.prod.outlook.com>
Comments: In-reply-to Murthy Nunna <mnunna@fnal.gov>
	message dated "Thu, 16 Jul 2020 21:10:00 -0000"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <3464717.1594936870.1@sss.pgh.pa.us>
Content-Transfer-Encoding: quoted-printable
Date: Thu, 16 Jul 2020 18:01:10 -0400
Message-ID: <3464718.1594936870@sss.pgh.pa.us>
Precedence: bulk

Murthy Nunna <mnunna@fnal.gov> writes:
> I have a question about replication... Can I replicate to a server runni=
ng different version of Linux OS (e.g Primary on RHEL 7 and stand-by on RH=
EL 6.1).

It's a little risky because of the possibility that the different OS
versions have different behavior for the "same" LC_COLLATE setting,
leading to different sort orders for text columns, which would make
indexes on text columns incorrect in the standby's worldview.

If you are only using "C" collation, or if you check that the collation(s)
you use behave the same on both platforms, it should be OK.

Also note that the hardware architecture had better be the same, eg
no 32-bit-to-64-bit cases.  But Postgres can detect that type of problem
for itself, generally speaking.  I emphasize collation hazards because
we lack any reliable detection for collation mismatch.

			regards, tom lane