Обсуждение: Skip recovery/standby signal files in pg_basebackup

Hackers,

Restoring these files could cause surprising behaviors so it seems best
to let the restore process create them when needed.

Patch is attached.

-- 
-David
david@pgmasters.net

On Sat, Sep 28, 2019 at 3:53 AM David Steele <david@pgmasters.net> wrote:
>
> Hackers,
>
> Restoring these files could cause surprising behaviors so it seems best
> to let the restore process create them when needed.

It's not a normal situation where a running postgres has either
recovery.signal or standby.signal but I'm +1 on this change for
safety.

The patch looks good to me.

Regards,

--
Masahiko Sawada

On Fri, Sep 27, 2019 at 02:52:54PM -0400, David Steele wrote:
> Restoring these files could cause surprising behaviors so it seems best
> to let the restore process create them when needed.
>
> Patch is attached.

When taking a base backup from a standby, we have always copied
recovery.conf if present, which would have triggered recovery (and
a standby if standby_mode was enabled).  Hence always including
RECOVERY_SIGNAL_FILE would be consistent with the past behavior.

Including STANDBY_SIGNAL_FILE would be consistent with checking if
standby_mode was set or not in recovery.conf.  We have replaced
standby_mode by the standby signal file, so including it if present
is consistent with the past as well, no?
--
Michael

Hi David,

On Sat, Sep 28, 2019 at 12:23 AM David Steele <david@pgmasters.net> wrote:
>
> Hackers,
>
> Restoring these files could cause surprising behaviors so it seems best
> to let the restore process create them when needed.
>

Could you please let us know what is the surprising behaviour you are
talking about here when including recovery/standby signal files in
pg_basebackup output.

If including recovery.conf in pg_basebackup output earlier wasn't a
problem then why including recovery/standby.signal should be a
problem.

Your patch is just trying to skip standby.signal or recovery.signal
files when the base backup is either taken on standby server or it is
taken on the server where the PITR is still going on or may be paused.

What would be the behaviour with your patch when *-R* option is used
with pg_basebackup to take backup from standby server ? Won't it
create a standby.signal file.

> Patch is attached.
>

--
With Regards,
Ashutosh Sharma
EnterpriseDB:http://www.enterprisedb.com