Re: Proposal for changes to recovery.conf API

On 29 November 2016 at 15:13, Simon Riggs <simon(at)2ndquadrant(dot)com> wrote:
> On 14 November 2016 at 15:50, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
>> On Sat, Nov 12, 2016 at 11:09 AM, Andres Freund <andres(at)anarazel(dot)de> wrote:
>>> I'm very tempted to rename this during the move to GUCs
>> ...
>>> Slightly less so, but still tempted to also rename these. They're not
>>> actually necessarily pointing towards a primary, and namespace-wise
>>> they're not grouped with recovery_*, which has become more important now
>>> that recovery.conf isn't a separate namespace anymore.
>>
>> -1 for renaming these. I don't think the current names are
>> particularly bad, and I think trying to agree on what would be better
>> could easily sink the whole patch.
>
> OK, so we can move forward. Thanks.
>
> I'm going to be doing final review and commit this week, at the
> Developer meeting on Thurs and on Friday, with input in person and on
> list.

err... no I'm not, based on review feedback in Tokyo.

New schedule is roughly this...
* agree changes over next week
* make changes and submit new patch by 1 Jan
* commit patch by 7 Jan

Overview of details agreed in Tokyo, now subject to further comments
from hackers

* Move recovery.conf parameters into postgresql.conf
Allow reload of most parameters, allow ALTER SYSTEM
Provide visibility of values through GUC interface

* recovery.conf is replaced by recovery.trigger -> recovery.done

* pg_basebackup -R
will write recovery.trigger to data directory
insert parameters postgresql.conf.auto, if possible

* backwards compatibility - read recovery.conf from $DATADIR -
presence of recovery.conf will cause ERROR
* backwards compatibility - some parameter names will change, so
allows others to changes also if needed

* Add docs: "Guide to changes in recovery.conf in 10.0"

* recovery_target as a single parameter, using proposed "xid 666"
"target value" format

* remove hot_standby parameter altogether, in line with earlier changes

* trigger_file renamed to promote_trigger_file

* standby_mode = on| off -> default to on

* pg_ctl recover - not as part of this patch

I'll work on the patch from here on, to allow me to work towards commit.

Comments please.

--
Simon Riggs http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services