MIME-Version: 1.0
From: Pratik Pandit <pratikpandit5322@gmail.com>
Date: Wed, 3 Jun 2026 23:50:27 +0530
Message-ID: 
 <CAFchsDX8Wt9Emwx1ESYSx0x+w7GZYLu=mjh3QfNRxcWe9kKR6A@mail.gmail.com>
Subject: Best practices for a 2-node Patroni + PostgreSQL cluster with a
 external/witness DCS
To: pgsql-admin@lists.postgresql.org
Content-Type: multipart/alternative; boundary="0000000000001951d006535d7c4a"
Archived-At: 
 <https://www.postgresql.org/message-id/CAFchsDX8Wt9Emwx1ESYSx0x%2Bw7GZYLu%3Dmjh3QfNRxcWe9kKR6A%40mail.gmail.com>
Precedence: bulk

--0000000000001951d006535d7c4a
Content-Type: text/plain; charset="UTF-8"

Hello PostgreSQL Community,

I am planning to set up a high-availability PostgreSQL cluster using
Patroni. Due to infrastructure constraints, I only have two dedicated
database nodes available.
I know that a Distributed Configuration Store (like etcd or Consul)
requires an odd number of nodes (minimum 3) to achieve quorum and safely
prevent split-brain scenarios.

Before I begin the deployment, I would appreciate your advice on the
following architectural approaches:
 1. Is it recommended to run a 3-node etcd cluster where 2 instances live
on the DB nodes and a 3rd "witness" instance lives on a lightweight
application server/VM?
2. Alternatively, does Patroni safely support a 2-node setup using a single
external DCS instance, or does that introduce a critical single point of
failure (SPOF)?

Any architectural pointers or template configurations for this constraint
would be highly appreciated.
Thanks!

--0000000000001951d006535d7c4a
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div dir=3D"ltr"><p>Hello PostgreSQL Community,</p><p>I am=
 planning to set up a high-availability PostgreSQL cluster using Patroni. D=
ue to infrastructure constraints, I only have two dedicated database nodes =
available.<br>I know that a Distributed Configuration Store (like etcd or C=
onsul) requires an odd number of nodes (minimum 3) to achieve quorum and sa=
fely prevent split-brain scenarios.<br><br>Before I begin the deployment, I=
 would appreciate your advice on the following architectural approaches:<br=
>=C2=A01. Is it recommended to run a 3-node etcd cluster where 2 instances =
live on the DB nodes and a 3rd &quot;witness&quot; instance lives on a ligh=
tweight application server/VM?<br>2. Alternatively, does Patroni safely sup=
port a 2-node setup using a single external DCS instance, or does that intr=
oduce a critical single point of failure (SPOF)?<br><br>Any architectural p=
ointers or template configurations for this constraint would be highly appr=
eciated.<br>Thanks!</p></div>
</div>

--0000000000001951d006535d7c4a--