Message-ID: <1513243810.5112.8.camel@cybertec.at>
Subject: Re: virtualized snapshots and PITR
From: Laurenz Albe <laurenz.albe@cybertec.at>
To: Achilleas Mantzios <achill@matrix.gatewaynet.com>,
	pgsql-admin@lists.postgresql.org
Date: Thu, 14 Dec 2017 10:30:10 +0100
In-Reply-To: <d9646cc6-167b-6422-f815-124d476623d5@matrix.gatewaynet.com>
References: <38c84302-270e-a2b6-648f-26ba798d8787@propaas.com>
	 <1513239540.2442.22.camel@cybertec.at>
	 <d9646cc6-167b-6422-f815-124d476623d5@matrix.gatewaynet.com>
Organization: Cybertec
Content-Type: text/plain; charset="UTF-8"
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Precedence: bulk

Achilleas Mantzios wrote:
> > > Anyone know of a way to do PITR recoveries with veeam?
> > 
> > It is fairly simple.
> > 
> > Before the snapshot, run
> > 
> >      SELECT pg_start_backup('label of your choice', TRUE);
> > 
> > After the snapshot, run
> > 
> >      SELECT pg_stop_backup();
> 
> This will create the data/backup_label file necessary for a minimal consistent recover,
> but wouldn't he also have a recovery.conf in place, so that the server will keep
> reading wals till it reaches the recovery target?

Yes, for point-in-time recovery you would need a recovery.conf file
with at least a "restore_command" in it.

> And if the system comes up live he won't have the chance to create this directory.

That's an interesting point; I didn't think of it.

One thing I can think of is to disable PostgreSQL autostart
while you take the snapshot.

> Basically what we do ourselves is exactly what you suggest.
> I'll have to ask if state (mem/cpu/registers/etc) is restored as well.

Such a kind of snapshot wouldn't be useful for a database backup.
You only want the storage.

Yours,
Laurenz Albe