At 2016-09-28 13:13:56 -0400, robertmhaas@gmail.com wrote:
>
> I hope that the fact that there's been no discussion for the last
> three weeks doesn't mean this effort is dead; I would like very
> much to see it move forward.
Here's an updated patch. Sorry, I got busy elswhere.
I struggled with the handling of recovery_target a little. For example,
one suggested alternative was:
recovery_target_type = xid
recovery_target_value = …
The problem with implementing it this way is that the _value setting
cannot be parsed without already having parsed the _type, and I didn't
want to force that sort of dependency.
What I've done instead is to make this work:
recovery_target = xid|time|name|lsn|immediate
recovery_target_xid = …
recovery_target_time = …
recovery_target_name = …
recovery_target_lsn = …
The recovery_target_xxx values are parsed as they used to be, but the
one that's used is the one that's set in recovery_target. That's easy to
explain, and the patch is much less intrusive, but I'm certainly open to
suggestions to improve this, and I have the time to work on this patch
with a view towards getting it committed in this cycle.
-- Abhijit
P.S. Sorry, I haven't been able to resolve the conflicts between Simon's
earlier recovery_startup_r10_api.v1b patch and the "pg_ctl promote -w"
changes in master. I was distracted by some illness in the family, but
I will post another update very soon.