Re: Add exclusive backup deprecation notes to documentation

David Steele <david@pgmasters.net>

From: David Steele <david@pgmasters.net>
To: Peter Eisentraut <peter.eisentraut@2ndquadrant.com>, Magnus Hagander <magnus@hagander.net>
Cc: Laurenz Albe <laurenz.albe@cybertec.at>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>, Martín Marqués <martin@2ndquadrant.com>, Michael Paquier <michael@paquier.xyz>
Date: 2019-03-29T12:33:29Z
Lists: pgsql-hackers
On 3/29/19 11:58 AM, Peter Eisentraut wrote:
> On 2019-03-20 14:42, Magnus Hagander wrote:
>> But that would be factually incorrect and backwards, so it seems like a
>> terrible idea, at least when it comes to manual. If you are doing it
>> manually, it's a lot *easier* to do it right with the non-exclusive
>> mode, because you can easily keep one psql and one shell open. And
>> that's safe.
> 
> The scenario I have in mind is, a poorly maintained server, nothing
> installed, can't install anything (no internet connection, license
> expired), flaky network, you fear it's going to fail soon, you need to
> take a backup.  The simplest procedure would appear to be: start backup
> mode, copy files away, stop backup mode.  Anything else that involves
> holding a session open over there for the whole time is way more fragile
> unless proper preparations have been made (and even then).  So I don't
> know what you want to call that scenario, but I would feel more
> comfortable having these basic tools available in a bind.

I would argue the best thing in this scenario is to use pg_basebackup. 
It's a solid tool and likely far better than any script the user might 
cook up on the spot.

Regards,
-- 
-David
david@pgmasters.net



Commits

  1. Warn more strongly about the dangers of exclusive backup mode.