Message-ID: <976875a5adf1c1cda0d55e6da0313f30dc8059f6.camel@cybertec.at>
Subject: Re: PostgreSQL replication failover
From: Laurenz Albe <laurenz.albe@cybertec.at>
To: Jan Peters <haseningo@gmx.de>, pgsql-admin@lists.postgresql.org
Date: Wed, 13 Jan 2021 12:25:35 +0100
In-Reply-To: 
 <trinity-64a56362-8af1-4ec9-b408-d4256ce65fd2-1610526479876@3c-app-gmx-bs17>
References: 
	<trinity-64a56362-8af1-4ec9-b408-d4256ce65fd2-1610526479876@3c-app-gmx-bs17>
Content-Type: text/plain; charset="UTF-8"
User-Agent: Evolution 3.36.5 (3.36.5-2.fc32) 
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Precedence: bulk

On Wed, 2021-01-13 at 09:27 +0100, Jan Peters wrote:
> we are running postgresqlserver on s390 zLinux machines. The distribution
>  is RedHat 7 and RedHat 8, so we do not have the many x86 tools available.
> 
> We always run 2 instances with a replication (streaming) async mode, the replica
>  is in hot_standby and we use it for read-only accesses. About the setup we have the following question:
> 
> How is an orderly failover accomplished? Our current procedure is.
> 
> 1. primary stop
> 2. promote replica to primary
> 3. create standby.signal on old primary
> 4. change primary_conninfo on old primary
> 5. start old primary as new replica
> 
> Is this processing correct? Are there any other steps that simplify a failover?
>  How can we be sure that all changes have been transferred from the old master to the replica?

What you describe is not a failover, but a switchover.

If you shut down the primary server cleanly, all changes will be replicated,
so you should be good.

During a failover, that is, if the primary suddenly fails, there is always
the possibility that you lose some transactions, unless you use synchronous
replication.

Yours,
Laurenz Albe
-- 
Cybertec | https://www.cybertec-postgresql.com