Re: pg_stop_backup(wait_for_archive := true) on standby server

On Tue, Jul 25, 2017 at 4:43 AM, Michael Paquier
<michael(dot)paquier(at)gmail(dot)com> wrote:
> On Mon, Jul 24, 2017 at 6:45 PM, Stephen Frost <sfrost(at)snowman(dot)net> wrote:
>> What the change would do is make the pg_stop_backup() caller block until
>> the last WAL is archvied, and perhaps that ends up taking hours, and
>> then the connection is dropped for whatever reason and the backup fails
>> where it otherwise.... what? wouldn't have been valid anyway at that
>> point, since it's not valid until the last WAL is actually archived.
>> Perhaps eventually it would be archived and the caller was planning for
>> that and everything is fine, but, well, that feels like an awful lot of
>> wishful thinking.
>
> Letting users taking unconsciously inconsistent backups is worse than
> potentially breaking scripts that were actually not working as
> Postgres would expect. So I am +1 for back-patching a lighter version
> of the proposed patch that makes the wait happen on purpose.
>
>>> > I'd hate to have to do it, but we could technically add a GUC to address
>>> > this in the back-branches, no? I'm not sure that's really worthwhile
>>> > though..
>>>
>>> That would be mighty ugly.
>>
>> Oh, absolutely agreed.
>
> Yes, let's avoid that. We are talking about a switch aimed at making
> backups potentially inconsistent.

Thank you for the review comments!
Attached updated the patch. The noting in pg_baseback doc will be
necessary for back branches if we decided to not back-patch it to back
branches. So it's not contained in this patch for now.

Regarding back-patching this to back branches, I also vote for
back-patching to back branches. Or we can fix the docs of back
branches and fix the code only in PG10. I expect that the user who
wrote a backup script has done enough functional test and dealt with
this issue somehow, but since the current doc clearly says that
pg_stop_backup() waits for all WAL to be archived we have to make a
consideration about there are users who wrote a wrong backup script.
So I think we should at least notify it in the minor release.

Regards,

--
Masahiko Sawada
NIPPON TELEGRAPH AND TELEPHONE CORPORATION
NTT Open Source Software Center