Обсуждение: Concrete steps for use of PITR warm backup?

I think my needs are standard for PITR: single master server and single
backup server, warm, with desire for PITR.

I'm running postgres 8.4.

I've looked at Postgres docs on warm backup, and several blogs
attempting instructions to help.

Everything I've seen does a good job of documenting the
initial set up:

- how to backup the postgres instance using rsync to second server
- how to enable wal archiving with rsync to second server

After that, the instructions are less clear.

I've also checked out barman.  It also is good at describing
the initial set up, but when it is time to describe PITR
use, they defer to Postgres docs and training which they sell.

Is there a good guide with examples on how to configure and
*use* a PITR solution?

I'm having problems with finding examples discussing:

- how to start up backup postgres and incorporate wal files
- how to maintain the build up of wal files (with explicit example)
- whether to run (or not run) the backup postgres service

I've made attempts to set up the backup, with failure.  It seems
we need the WAL files in the pg_xlog directory, but postgres
refuses to start.  It isn't surprising to me as I had to try
rather than know.

Postgres documentation is written with too many statements along the
"If... Then..." format.  When you already know what to do, this
format is sufficient.  When you are looking for a set of steps,
even an example scenario, this is insufficient.

I'd think the scenario of single master, single backup, PITR with warm
recovery from WAL archive, is very typical and what many people
are looking for.  Can't it be documented clearly with
a complete set of steps?

On Thu, Jun 13, 2013 at 10:51 AM, francis picabia <fpicabia@gmail.com> wrote:
> I think my needs are standard for PITR: single master server and single
> backup server, warm, with desire for PITR.
>
> I'm running postgres 8.4.
>
> I've looked at Postgres docs on warm backup, and several blogs
> attempting instructions to help.
>
> Everything I've seen does a good job of documenting the
> initial set up:
>
> - how to backup the postgres instance using rsync to second server
> - how to enable wal archiving with rsync to second server
>
> After that, the instructions are less clear.
>
> I've also checked out barman.  It also is good at describing
> the initial set up, but when it is time to describe PITR
> use, they defer to Postgres docs and training which they sell.
>
> Is there a good guide with examples on how to configure and
> *use* a PITR solution?
>
> I'm having problems with finding examples discussing:
>
> - how to start up backup postgres and incorporate wal files
> - how to maintain the build up of wal files (with explicit example)
> - whether to run (or not run) the backup postgres service
>
> I've made attempts to set up the backup, with failure.  It seems
> we need the WAL files in the pg_xlog directory, but postgres
> refuses to start.  It isn't surprising to me as I had to try
> rather than know.
>
> Postgres documentation is written with too many statements along the
> "If... Then..." format.  When you already know what to do, this
> format is sufficient.  When you are looking for a set of steps,
> even an example scenario, this is insufficient.
>
> I'd think the scenario of single master, single backup, PITR with warm
> recovery from WAL archive, is very typical and what many people
> are looking for.  Can't it be documented clearly with
> a complete set of steps?

I still think the documentation is lacking real life examples, but my
needs have been met by pgbarman.  It is very good, saves developing
a script yourself, and works at 4 AM when your brain is not.  I also
like that barman is so easy that I can go on vacation and anyone
can follow the instructions for recovery.