On Sat, Aug 04, 2018 at 07:44:59AM +0100, Simon Riggs wrote:
> I think the problem is that writing the online checkpoint is deferred
> after promotion, so this is a timing issue that probably doesn't show
> in our regression tests.
Somewhat. It is a performance improvement of 9.3 to let the startup
request a checkpoint to the checkpointer process instead of doing it
itself.
> Sounds like we should write a pending timeline change to the control
> file and have pg_rewind check that instead.
>
> I'd call this a timing bug, not a doc issue.
Well, having pg_rewind enforce a checkpoint on the promoted standby
could cause a performance hit as well if we do it mandatorily as if
there is delay between the promotion and the rewind triggerring a
checkpoint could have already happen. So it is for me a documentation
bug first regarding the failover workflow, and potentially a patch for a
new feature which makes pg_rewind trigger directly a checkpoint.
--
Michael