On Fri, Feb 15, 2019 at 08:06:16AM +0900, Michael Paquier wrote: > Thanks, I see what you have done. I cannot comment if your shortcut > is actually fully correct based on my knowledge of this code, but > things cannot be in the best conditions without having the WAL reader > handle properly the limits. So I am planning to commit what you > propose after an extra pass on it in the next couple of days or so.
And done, after doing and extra pass, doing more testing using by own plugins, pg_waldump and more fancy stuff with a primary/standby and pgbench. -- Michael