Re: pg_dumpall with individual output files per database? - Mailing list pgsql-general

From Francisco Olarte
Subject Re: pg_dumpall with individual output files per database?
Date
Msg-id CA+bJJbwxTJMqoVaYss_tE_pw2A=+STD25ORE1_-O9_Vj_0FYow@mail.gmail.com
Whole thread Raw
In response to pg_dumpall with individual output files per database?  (Thorsten Schöning<tschoening@am-soft.de>)
Responses Re: pg_dumpall with individual output files per database?
List pgsql-general
Hi Thorsten:

On Thu, Jul 15, 2021 at 6:30 PM Thorsten Schöning <tschoening@am-soft.de> wrote:

> I need to backup multiple host with multiple Postgres databases each.
> In all of those cases I'm interested to backup all databases, which
> makes pg_dumpall a natural choice at first glance. Though, from my
> understanding of the docs that is only capable of storing all
> databases per host into one single file.

I think what you are naming host is better named
server/cluster/instance ( a host may be running several postgres
installations, pg_dumpall dos not dump the wole host just the instance
you point it too ).

...
> So, is there some option I'm missing telling pg_dumpall to dump into
> individual files, simply named after e.g. the dumped databases?
> If not, was a feature like that discussed already or what's the
> reasons to not do that? There are a lot of search results how to dump
> all databases with lots of different scripting approaches. Many of
> those could simply be avoided with pg_dumpall supporting that
> already.

It would probably complicate it, and dumping a whole cluster using
something like pg_dumpall -g for the globals plus a loop over the
databases using something like pg_dump -Fc ( which I would always
recommend over plain sql format ) is just a ( little complex ) one
liner or a 10 line script, probably not worth the scarce developer /
maintainer brain cycles.

> Tools like BorgMatic making use of pg_dumpall might benefit of such a
> feature as well:

They might, but in a project of the (apparent, have not dug out much )
size of that, I would possibly just include an script, or dump the
databases as individual backup objects ( different retention cycles /
copies per database, skipping of dev/test databases etc). The script
to dump a whole cluster is just one line for the pg_dumpall, one psql
line to grab the database names and another for a loop pg_dumping all
of them. And from them on you can improve it a bit for special client
software purposes, it does not seem to hold its weight.

FOS



pgsql-general by date:

Previous
From: Thorsten Schöning
Date:
Subject: pg_dumpall with individual output files per database?
Next
From: Lucas
Date:
Subject: Re: PostgreSQL 9.2 high replication lag